技术领域Technical Field
本发明涉及智能看护领域，更具体地说，涉及一种基于IoT的智能养老看护系统及看护方法。The present invention relates to the field of intelligent nursing, and more specifically, to an IoT-based intelligent elderly care system and a nursing method.
背景技术Background Art
目前，随着社会老龄化的趋势日益严重，独居老人的安全和健康问题受到了广泛关注。由此，智能看护系统已经成为一种重要的解决方案。然而，现有的智能看护系统在实际应用中存在一些明显的不足。At present, with the increasingly serious trend of social aging, the safety and health of elderly people living alone have received widespread attention. Therefore, the intelligent care system has become an important solution. However, the existing intelligent care system has some obvious shortcomings in practical applications.
首先，大多数智能看护系统的设计和功能主要基于通用的生理数据和行为模式，缺乏对个体差异的深入理解和有效适应。这意味着它们可能无法充分地满足每个独居老人的具体需求和生活习惯，从而影响了看护的效果和用户的满意度。First, the design and functions of most smart care systems are mainly based on general physiological data and behavioral patterns, lacking in-depth understanding and effective adaptation to individual differences. This means that they may not fully meet the specific needs and living habits of each elderly person living alone, thus affecting the effectiveness of care and user satisfaction.
其次，虽然一些看护系统已经开始尝试使用语音交互功能，以便更好地与老人沟通，但这些系统往往只是用于简单的语音交流，无法结合对于老人生活习惯的了解来对于老人行为状态进行智能地纠正和提醒。Secondly, although some care systems have begun to try to use voice interaction functions to better communicate with the elderly, these systems are often only used for simple voice communication and cannot combine understanding of the elderly’s living habits to intelligently correct and remind the elderly’s behavioral status.
另外，当前的看护系统大多数是基于静态的模型和规则，无法自我学习和自我改进。这意味着，当遇到新的情况或者环境变化时，这些系统可能无法做出正确的响应。更重要的是，这些系统无法从每一次的看护经验中学习，进而提升其看护的效果和质量；In addition, most current care systems are based on static models and rules and are unable to self-learn and self-improve. This means that when encountering new situations or changes in the environment, these systems may not be able to respond correctly. More importantly, these systems cannot learn from each care experience and thus improve the effectiveness and quality of their care;
最后，当前老人看护系统主要关注老人的生理健康状况，而忽略了老人的情绪、心理健康，而情绪心理健康实际上也是老人的整体健康的非常大的一部分，且其往往会影响到生理健康状态。Finally, the current elderly care system mainly focuses on the physical health of the elderly, while ignoring their emotional and mental health. Emotional and mental health is actually a very large part of the overall health of the elderly, and it often affects their physical health.
因此，为了解决这些问题，急需一种新的智能看护系统，它不仅能够进行个性化的看护，提供准确的语音交互，而且能够针对老人的情绪进行特别提醒、关照和看护，其可实现增量学习，以提升看护的效果和用户的满意度。Therefore, in order to solve these problems, a new intelligent care system is urgently needed, which can not only provide personalized care and accurate voice interaction, but also provide special reminders, care and care based on the emotions of the elderly. It can realize incremental learning to improve the effect of care and user satisfaction.
发明内容Summary of the invention
本发明要解决的技术问题是提供一种基于IoT的智能养老看护系统及看护方法，以解决背景技术中提到的问题。The technical problem to be solved by the present invention is to provide an IoT-based intelligent elderly care system and a care method to solve the problems mentioned in the background technology.
为了达到上述目的，本发明采取以下技术方案：In order to achieve the above object, the present invention adopts the following technical solutions:
一种基于IoT的智能养老看护系统，包括传感器模块、数据处理与控制模块、报警模块和交流模块；An IoT-based intelligent elderly care system, comprising a sensor module, a data processing and control module, an alarm module and a communication module;
传感器模块位于老人的生活区域内，其中包括高精度毫米波雷达传感器和语音传感器；传感器模块还包括老人随身携带的生物传感器，实时获取老人的各项生理参数；The sensor module is located in the elderly's living area, including a high-precision millimeter-wave radar sensor and a voice sensor; the sensor module also includes a biosensor that the elderly carry with them to obtain their various physiological parameters in real time;
数据处理与控制模块与传感器模块连接，用于接收和处理传感器模块提供的数据；数据处理与控制模块利用医疗大数据和历史医疗数据，结合实时生理数据和情绪数据的变化，预测老人的行为并生成预警预测信息；数据处理与控制模块连接报警模块和提醒模块；The data processing and control module is connected to the sensor module to receive and process the data provided by the sensor module; the data processing and control module uses medical big data and historical medical data, combined with changes in real-time physiological data and emotional data, to predict the behavior of the elderly and generate early warning prediction information; the data processing and control module is connected to the alarm module and the reminder module;
报警模块通过内置的无线通信装置与老人家属或救护中心进行通信；The alarm module communicates with the elderly's family or rescue center through a built-in wireless communication device;
交流模块包括设置在老人生活区域的音箱或者老人随身佩戴的语音装置；The communication module includes a speaker installed in the elderly's living area or a voice device that the elderly wear;
所述传感器模块获取老人的语音并进行识别，数据处理与控制模块根据语音识别老人的情绪，对获取的情绪数据进行情绪分析并依据分析结果做个性化音频播放，通过交流模块播放声音或音乐来改善老人的情绪，声音的音色根据老人家人的声音进行特别设置；The sensor module acquires the elderly's voice and recognizes it, the data processing and control module recognizes the elderly's emotions based on the voice, performs emotion analysis on the acquired emotion data and performs personalized audio playback based on the analysis results, and plays sound or music through the communication module to improve the elderly's mood, and the timbre of the sound is specially set according to the voice of the elderly's family members;
数据处理与控制模块结合预警预测信息，启动辅助设备对风险老人进行重点看护，包括情感对话、提醒、辅助机器人或智能座椅；The data processing and control module combines the early warning and prediction information to activate auxiliary equipment to provide key care for the elderly at risk, including emotional dialogue, reminders, auxiliary robots or smart seats;
其中所述情绪分析使用卷积神经网络CNN和长短期记忆网络LSTM的结合来进行处理，具体包括：The sentiment analysis is processed by combining the convolutional neural network (CNN) and the long short-term memory (LSTM) network, specifically including:
将老人语音数据转化为频谱图像或梅尔频率倒谱系数MFCC，使用一维卷积处理MFCC特征，以提取局部的频率模式：Hc＝CNN(X)；其中，Hc是卷积层的输出，CNN(X)代表对输入的MFCC特征X进行卷积操作；The elderly speech data is converted into a spectrum image or Mel-frequency cepstral coefficient MFCC, and the MFCC features are processed using one-dimensional convolution to extract local frequency patterns: Hc = CNN(X); where Hc is the output of the convolution layer, and CNN(X) represents the convolution operation on the input MFCC feature X;
然后用LSTM处理卷积层的输出，以捕捉长期的时间依赖性：Hl＝LSTM(Hc)；其中，Hl是LSTM层的输出；The output of the convolutional layer is then processed with LSTM to capture long-term temporal dependencies: Hl = LSTM(Hc); where Hl is the output of the LSTM layer;
最后，用一个全连接层进行情绪分类：Y＝softmax(W×Hl+b)；其中，W和b是全连接层的权重和偏置，softmax函数用来输出每个情绪类别的概率，Y代表预测的情绪类别；Finally, a fully connected layer is used for emotion classification: Y = softmax(W×Hl+b); where W and b are the weights and biases of the fully connected layer, the softmax function is used to output the probability of each emotion category, and Y represents the predicted emotion category;
对应的情绪类比以及交流模块的回应包括：The corresponding emotional analogies and responses of the communication module include:
快乐:当老人处于快乐的情绪状态时，交流模块播放欢快的音乐，包括流行音乐或经典老歌；或使用温暖，友好的语音回应；Happiness: When the elderly are in a happy mood, the communication module plays cheerful music, including pop music or classic old songs; or responds with a warm, friendly voice;
悲伤:如果老人的情绪状态被识别为悲伤，交流模块播放舒缓的、缓解压力的音乐，包括古典音乐或轻音乐；或使用温和、安慰的语音交互；Sadness: If the elderly person’s emotional state is identified as sadness, the communication module plays soothing, stress-relieving music, including classical music or light music; or uses gentle, comforting voice interaction;
焦虑或恐惧:当老人处于焦虑或恐惧的情绪状态时，播放自然的声音或冥想音乐以缓解焦虑；交流模块的语音保持平稳，镇定，并提供保证或安慰的语言；Anxiety or fear: When the elderly are in an anxious or fearful emotional state, play natural sounds or meditation music to relieve anxiety; the voice of the communication module remains smooth and calm, and provides words of reassurance or comfort;
愤怒:如果检测到老人的情绪状态是愤怒，可以播放平静的音乐，如爵士乐或蓝调；同时，交流模块使用平稳、理解的语音，避免挑衅或惹怒老人；Anger: If the elderly person's emotional state is detected to be anger, calming music such as jazz or blues can be played; at the same time, the communication module uses a smooth, understanding voice to avoid provoking or angering the elderly person;
中性:当老人处于中性情绪状态时，可以播放他们喜欢的音乐；交流模块可以保持中性、友好的语音；Neutral: When the elderly are in a neutral emotional state, their favorite music can be played; the communication module can maintain a neutral and friendly voice;
所述系统还包括如下工作流程：The system also includes the following workflow:
S1：所述传感器模块将收集到的有关老人的传感器数据及对应的时间发送到数据处理与控制模块；S1: The sensor module sends the collected sensor data about the elderly and the corresponding time to the data processing and control module;
S2：每当数据处理与控制模块识别到传感器数据的变化速率超过预设阈值时，通过交流模块询问老人已完成的行为以及其该行为的健康程度，并结合询问结果以及已接收到的传感器数据，获取传感器数据与老人行为的关联关系，并同时标记该行为的发生时间段以及对应的健康程度；当数据处理与控制模块的所接收到的数据量积累到预设数据量时，进入步骤S3～S5；S2: Whenever the data processing and control module identifies that the rate of change of the sensor data exceeds a preset threshold, the communication module asks the elderly about the completed behavior and the health level of the behavior, and combines the inquiry result with the received sensor data to obtain the correlation between the sensor data and the elderly's behavior, and simultaneously marks the time period of the behavior and the corresponding health level; when the amount of data received by the data processing and control module accumulates to the preset amount of data, it enters steps S3 to S5;
S3：数据处理与控制模块对老人的所有行为及其对应的发生时间进行关联得到一系列行为序列作为老人的总体行为模式；S3: The data processing and control module associates all behaviors of the elderly and their corresponding occurrence times to obtain a series of behavior sequences as the overall behavior pattern of the elderly;
S4：数据处理与控制模块对老人的超过健康阈值的所有行为及其对应的发生时间进行关联得到一系列行为序列作为老人的健康行为模式；S4: The data processing and control module associates all behaviors of the elderly that exceed the health threshold and their corresponding occurrence time to obtain a series of behavior sequences as the healthy behavior pattern of the elderly;
S5：数据处理与控制模块对老人的低于健康阈值的所有行为及其对应的发生时间进行关联得到一系列行为序列作为老人的非健康行为模式；S5: The data processing and control module associates all behaviors of the elderly below the health threshold and their corresponding occurrence times to obtain a series of behavior sequences as the elderly's unhealthy behavior patterns;
S6：当数据处理与控制模块得到老人的总体行为模式、健康行为模式和非健康行为模式时，通过交流模块提示老人可进入看护阶段；S6: When the data processing and control module obtains the overall behavior pattern, healthy behavior pattern and unhealthy behavior pattern of the elderly, the communication module prompts the elderly to enter the care stage;
S7：在看护阶段中，数据处理与控制模块实时获取传感器数据，并结合S2中的关联关系识别传感器数据对应的行为，并取预先设置的滑动窗口时间段内的随时间变化的行为序列与总体行为模式、健康行为模式以及非健康行为模式进行匹配，获取三个相似度，并将三个相似度中最高的相似度与预设相似度阈值进行比对；S7: In the care phase, the data processing and control module acquires sensor data in real time, identifies the behavior corresponding to the sensor data in combination with the association relationship in S2, and matches the behavior sequence that changes over time within a preset sliding window period with the overall behavior pattern, the healthy behavior pattern, and the unhealthy behavior pattern, obtains three similarities, and compares the highest similarity among the three similarities with the preset similarity threshold;
若三个相似度中，对应于健康行为模式的相似度最高且超过预设的匹配阈值，则数据处理与控制模块通过交流模块鼓励老人；If, among the three similarities, the similarity corresponding to the healthy behavior pattern is the highest and exceeds the preset matching threshold, the data processing and control module encourages the elderly through the communication module;
若三个相似度中，对应于非健康行为模式的相似度最高且超过预设的匹配阈值，则数据处理与控制模块通过交流模块提示老人要注意行为健康、播报老人在该滑动窗口内的所有行为，并依据健康行为模式对老人行为的预测并给予老人实时提醒、生活运动建议和饮食建议；If the similarity corresponding to the unhealthy behavior pattern is the highest among the three similarities and exceeds the preset matching threshold, the data processing and control module will remind the elderly to pay attention to behavioral health through the communication module, broadcast all the elderly's behaviors within the sliding window, and predict the elderly's behavior based on the healthy behavior pattern and give the elderly real-time reminders, life exercise suggestions and diet suggestions;
若三个相似度均不超过预设相似度阈值，则通过提示模块询问老人在滑动窗口内进行了哪些行为，若未得到老人的回应，则数据处理与控制模块通过报警模块与老人家属或救护中心进行报警；若得到老人回应，则数据处理与控制模块将该滑动窗口内的传感器数据及对应的行为、行为发生时间加入至步骤S2～S5中所采用的训练数据，进而采用增量学习的方式实时调整S2中的关联关系以及S3～S5中的三种行为模式，并将其实时应用至老人的看护阶段。If the three similarities do not exceed the preset similarity threshold, the prompt module will ask the elderly what behaviors they have performed in the sliding window. If there is no response from the elderly, the data processing and control module will alert the elderly's family or the rescue center through the alarm module; if the elderly responds, the data processing and control module will add the sensor data and the corresponding behaviors and the time when the behaviors occurred in the sliding window to the training data used in steps S2 to S5, and then use incremental learning to adjust the association relationship in S2 and the three behavior patterns in S3 to S5 in real time, and apply them in real time to the care stage of the elderly.
相似度可能通过一种称为时间序列相似度度量的方法计算。这种度量方法可以计算两个时间序列之间的相似性。有多种方法可以做到这一点，如动态时间弯曲(DTW),欧几里得距离(Euclidean Distance)或余弦相似性(Cosine Similarity)。让我们以DTW为例，因为它可以处理不同长度和速度的时间序列。Similarity may be calculated using a method called time series similarity measure. This measure calculates the similarity between two time series. There are various methods to do this, such as Dynamic Time Warping (DTW), Euclidean Distance or Cosine Similarity. Let's take DTW as an example, because it can handle time series of different lengths and speeds.
假设我们有两个时间序列Q和C，它们分别表示预测的行为模式和实时采集的行为模式，那么它们之间的DTW距离可以通过以下步骤计算：Assume that we have two time series Q and C, which represent the predicted behavior pattern and the real-time collected behavior pattern respectively, then the DTW distance between them can be calculated by the following steps:
初始化一个大小为len(Q)×len(C)的矩阵D，其中D[i][j]代表Q的前i个元素和C的前j个元素之间的DTW距离。Initialize a matrix D of size len(Q)×len(C), where D[i][j] represents the DTW distance between the first i elements of Q and the first j elements of C.
对于i＝1到len(Q)，j＝1到len(C)：For i=1 to len(Q), j=1 to len(C):
a.计算Q[i]和C[j]之间的距离：d＝|Q[i]-C[j]|。a. Calculate the distance between Q[i] and C[j]: d = |Q[i] - C[j]|.
b.更新D[i][j]＝d+min(D[i-1][j],D[i][j-1],D[i-1][j-1])。b. Update D[i][j]=d+min(D[i-1][j], D[i][j-1], D[i-1][j-1]).
返回D[len(Q)][len(C)]作为Q和C之间的DTW距离。Return D[len(Q)][len(C)] as the DTW distance between Q and C.
然后，可以通过以下公式将DTW距离转化为相似度：Then, the DTW distance can be converted into similarity by the following formula:
相似度＝exp(-DTW距离/阈值)；Similarity = exp(-DTW distance/threshold);
在这里，阈值是一个可以调整的参数，用来控制DTW距离和相似度之间的映射关系。Here, the threshold is an adjustable parameter used to control the mapping relationship between DTW distance and similarity.
最后，可以将计算出的相似度与预设的相似度阈值进行比较，以决定如何通过交流模块给老人反馈。Finally, the calculated similarity can be compared with the preset similarity threshold to decide how to give feedback to the elderly through the communication module.
优选的，所述健康程度通过健康程度因子评估，所述健康程度因子取0～100，数值越高表示行为越健康。Preferably, the health level is evaluated by a health level factor, and the health level factor ranges from 0 to 100, with a higher value indicating healthier behavior.
优选的，所述数据处理与控制模块包括：Preferably, the data processing and control module includes:
数据接收器，用于接收来自传感器模块的数据；A data receiver, used for receiving data from the sensor module;
数据处理系统，包括连接于数据接收器并用于存储传感器数据的数据库，以及用于处理数据的数据分析引擎；a data processing system including a database connected to the data receiver and used to store the sensor data, and a data analysis engine for processing the data;
决策引擎，用于根据数据分析结果做出决定并发送至报警模块或交流模块。The decision engine is used to make decisions based on the data analysis results and send them to the alarm module or communication module.
优选的，所述数据处理与控制模块运行在一个服务器上，所述服务器位于云中或用户的家庭网络中。Preferably, the data processing and control module runs on a server, which is located in the cloud or in the user's home network.
本发明还公开一种基于IoT的智能养老看护方法，包括如下步骤：The present invention also discloses an IoT-based intelligent elderly care method, comprising the following steps:
通过传感器模块获取老人的语音数据，将老人语音数据转化为频谱图像或梅尔频率倒谱系数MFCC，使用一维卷积处理MFCC特征，以提取局部的频率模式：Hc＝CNN(X)；其中，Hc是卷积层的输出，CNN(X)代表对输入的MFCC特征X进行卷积操作；The voice data of the elderly is obtained through the sensor module, and the voice data of the elderly is converted into a spectrum image or Mel frequency cepstral coefficient MFCC. The MFCC features are processed using one-dimensional convolution to extract the local frequency pattern: Hc = CNN(X); where Hc is the output of the convolution layer, and CNN(X) represents the convolution operation on the input MFCC feature X;
然后用LSTM处理卷积层的输出，以捕捉长期的时间依赖性：Hl＝LSTM(Hc)；其中，Hl是LSTM层的输出；The output of the convolutional layer is then processed with LSTM to capture long-term temporal dependencies: Hl = LSTM(Hc); where Hl is the output of the LSTM layer;
最后，用一个全连接层进行情绪分类：Y＝softmax(W×Hl+b)；其中，W和b是全连接层的权重和偏置，softmax函数用来输出每个情绪类别的概率，Y代表预测的情绪类别；Finally, a fully connected layer is used for emotion classification: Y = softmax(W×Hl+b); where W and b are the weights and biases of the fully connected layer, the softmax function is used to output the probability of each emotion category, and Y represents the predicted emotion category;
对应的情绪类比以及交流模块的回应包括：The corresponding emotional analogies and responses of the communication module include:
快乐:当老人处于快乐的情绪状态时，交流模块播放欢快的音乐，包括流行音乐或经典老歌；或使用温暖，友好的语音回应；Happiness: When the elderly are in a happy mood, the communication module plays cheerful music, including pop music or classic old songs; or responds with a warm, friendly voice;
悲伤:如果老人的情绪状态被识别为悲伤，交流模块播放舒缓的、缓解压力的音乐，包括古典音乐或轻音乐；或使用温和、安慰的语音交互；Sadness: If the elderly person’s emotional state is identified as sadness, the communication module plays soothing, stress-relieving music, including classical music or light music; or uses gentle, comforting voice interaction;
焦虑或恐惧:当老人处于焦虑或恐惧的情绪状态时，播放自然的声音或冥想音乐以缓解焦虑；交流模块的语音保持平稳，镇定，并提供保证或安慰的语言；Anxiety or fear: When the elderly are in an anxious or fearful emotional state, play natural sounds or meditation music to relieve anxiety; the voice of the communication module remains smooth and calm, and provides words of reassurance or comfort;
愤怒:如果检测到老人的情绪状态是愤怒，可以播放平静的音乐，如爵士乐或蓝调；同时，交流模块使用平稳、理解的语音，避免挑衅或惹怒老人；Anger: If the elderly person's emotional state is detected to be anger, calming music such as jazz or blues can be played; at the same time, the communication module uses a smooth, understanding voice to avoid provoking or angering the elderly person;
中性:当老人处于中性情绪状态时，可以播放他们喜欢的音乐；交流模块可以保持中性、友好的语音。Neutral: When the elderly are in a neutral emotional state, their favorite music can be played; the communication module can maintain a neutral and friendly voice.
优选的，所述方法还包括并行的如下步骤：Preferably, the method further comprises the following steps in parallel:
S1：所述传感器模块将收集到的有关老人的传感器数据及对应的时间发送到数据处理与控制模块；S1: The sensor module sends the collected sensor data about the elderly and the corresponding time to the data processing and control module;
S2：每当数据处理与控制模块识别到传感器数据的变化速率超过预设阈值时，通过交流模块询问老人已完成的行为以及其该行为的健康程度，并结合询问结果以及已接收到的传感器数据，获取传感器数据与老人行为的关联关系，并同时标记该行为的发生时间段以及对应的健康程度；当数据处理与控制模块的所接收到的数据量积累到预设数据量时，进入步骤S3～S5；S2: Whenever the data processing and control module identifies that the rate of change of the sensor data exceeds a preset threshold, the communication module asks the elderly about the completed behavior and the health level of the behavior, and combines the inquiry result with the received sensor data to obtain the correlation between the sensor data and the elderly's behavior, and simultaneously marks the time period of the behavior and the corresponding health level; when the amount of data received by the data processing and control module accumulates to the preset amount of data, it enters steps S3 to S5;
S3：数据处理与控制模块对老人的所有行为及其对应的发生时间进行关联得到一系列行为序列作为老人的总体行为模式；S3: The data processing and control module associates all behaviors of the elderly and their corresponding occurrence times to obtain a series of behavior sequences as the overall behavior pattern of the elderly;
S4：数据处理与控制模块对老人的超过健康阈值的所有行为及其对应的发生时间进行关联得到一系列行为序列作为老人的健康行为模式；S4: The data processing and control module associates all behaviors of the elderly that exceed the health threshold and their corresponding occurrence time to obtain a series of behavior sequences as the healthy behavior pattern of the elderly;
S5：数据处理与控制模块对老人的低于健康阈值的所有行为及其对应的发生时间进行关联得到一系列行为序列作为老人的非健康行为模式；S5: The data processing and control module associates all behaviors of the elderly below the health threshold and their corresponding occurrence times to obtain a series of behavior sequences as the elderly's unhealthy behavior patterns;
S6：当数据处理与控制模块得到老人的总体行为模式、健康行为模式和非健康行为模式时，通过交流模块提示老人可进入看护阶段；S6: When the data processing and control module obtains the overall behavior pattern, healthy behavior pattern and unhealthy behavior pattern of the elderly, the communication module prompts the elderly to enter the care stage;
S7：在看护阶段中，数据处理与控制模块实时获取传感器数据，并结合S2中的关联关系识别传感器数据对应的行为，并取预先设置的滑动窗口时间段内的随时间变化的行为序列与总体行为模式、健康行为模式以及非健康行为模式进行匹配，获取三个相似度，并将三个相似度中最高的相似度与预设相似度阈值进行比对；S7: In the care phase, the data processing and control module acquires sensor data in real time, identifies the behavior corresponding to the sensor data in combination with the association relationship in S2, and matches the behavior sequence that changes over time within a preset sliding window period with the overall behavior pattern, the healthy behavior pattern, and the unhealthy behavior pattern, obtains three similarities, and compares the highest similarity among the three similarities with the preset similarity threshold;
若三个相似度中，对应于健康行为模式的相似度最高且超过预设的匹配阈值，则数据处理与控制模块通过交流模块鼓励老人；If, among the three similarities, the similarity corresponding to the healthy behavior pattern is the highest and exceeds the preset matching threshold, the data processing and control module encourages the elderly through the communication module;
若三个相似度中，对应于非健康行为模式的相似度最高且超过预设的匹配阈值，则数据处理与控制模块通过交流模块提示老人要注意行为健康、播报老人在该滑动窗口内的所有行为，并依据健康行为模式对老人行为的预测给予老人实时提醒、生活运动建议和饮食建议；If the similarity corresponding to the unhealthy behavior pattern is the highest among the three similarities and exceeds the preset matching threshold, the data processing and control module will remind the elderly to pay attention to behavioral health through the communication module, broadcast all the elderly's behaviors within the sliding window, and give the elderly real-time reminders, life exercise suggestions and diet suggestions based on the prediction of the healthy behavior pattern;
若三个相似度均不超过预设相似度阈值，则通过提示模块询问老人在滑动窗口内进行了哪些行为，若未得到老人的回应，则数据处理与控制模块通过报警模块与老人家属或救护中心进行报警；若得到老人回应，则数据处理与控制模块将该滑动窗口内的传感器数据及对应的行为、行为发生时间加入至步骤S2～S5中所采用的训练数据，进而采用增量学习的方式实时调整S2中的关联关系以及S3～S5中的三种行为模式，并将其实时应用至老人的看护阶段。If the three similarities do not exceed the preset similarity threshold, the prompt module will ask the elderly what behaviors they have performed in the sliding window. If there is no response from the elderly, the data processing and control module will alert the elderly's family or the rescue center through the alarm module; if the elderly responds, the data processing and control module will add the sensor data and the corresponding behaviors and the time when the behaviors occurred in the sliding window to the training data used in steps S2 to S5, and then use incremental learning to adjust the association relationship in S2 and the three behavior patterns in S3 to S5 in real time, and apply them in real time to the care stage of the elderly.
优选的，S2中的关联关系采用监督学习中的分类算法实现，所述分类算法采用以下任意一种：决策树、随机森林、神经网络算法。Preferably, the association relationship in S2 is implemented using a classification algorithm in supervised learning, and the classification algorithm uses any one of the following: decision tree, random forest, neural network algorithm.
优选的，所述相似度通过动态时间规整算法进行计算。Preferably, the similarity is calculated by a dynamic time warping algorithm.
优选的，使用在线支持向量机作为增量学习算法。Preferably, an online support vector machine is used as the incremental learning algorithm.
本发明相对于现有技术的优点在于：The advantages of the present invention over the prior art are:
个性化看护：本发明通过实时收集和分析独居老人的行为和语音数据，能够深入理解每个老人的具体需求和生活习惯，从而提供个性化的看护服务。此外，本发明还能够实时监测老人的行为状态，并与老人进行语音交互，以提醒他们注意健康问题。Personalized care: The present invention can deeply understand the specific needs and living habits of each elderly person by collecting and analyzing their behavior and voice data in real time, thus providing personalized care services. In addition, the present invention can also monitor the behavior of the elderly in real time and interact with them by voice to remind them to pay attention to health problems.
智能语音交互：本发明不仅能够识别老人的语音信息，还能够通过语音反馈提供针对性的健康建议和提醒。这意味着，本发明能够根据老人的生活习惯和行为模式，智能地纠正和提醒老人的行为状态，从而帮助他们维持良好的生活习惯和健康状态。Intelligent voice interaction: The present invention can not only recognize the voice information of the elderly, but also provide targeted health advice and reminders through voice feedback. This means that the present invention can intelligently correct and remind the elderly's behavior status according to their living habits and behavior patterns, thereby helping them maintain good living habits and health status.
增量学习：本发明采用增量学习的方式，可以实时收集和学习新的数据，并根据新的数据调整看护模式。这意味着，本发明能够随着时间的推移，不断学习和改进，从而提高看护的效果和质量。Incremental learning: The present invention adopts an incremental learning approach, which can collect and learn new data in real time and adjust the care mode according to the new data. This means that the present invention can continuously learn and improve over time, thereby improving the effectiveness and quality of care.
即时报警：当老人的行为模式出现异常且无法得到老人的回应时，本发明的系统能够立即通过报警模块与老人的家属或救护中心进行通信，以确保老人的安全。Instant alarm: When the behavior pattern of the elderly is abnormal and the elderly cannot respond, the system of the present invention can immediately communicate with the elderly's family or rescue center through the alarm module to ensure the safety of the elderly.
动态调整：通过增量学习和实时匹配，本发明能够动态地调整老人的行为模式和健康阈值，以适应老人的生活习惯和健康状态的变化，从而提供更为精准和个性化的看护服务。Dynamic adjustment: Through incremental learning and real-time matching, the present invention can dynamically adjust the behavior patterns and health thresholds of the elderly to adapt to changes in their living habits and health status, thereby providing more accurate and personalized care services.
除此之外，本发明重点在于能够进行情绪识别，其有益效果在于：In addition, the present invention focuses on being able to perform emotion recognition, and its beneficial effects are:
引入情绪识别功能，可以更全面、更深度地理解和关注老人的健康状况。除了生理健康和行为模式外，现在还能够关注到老人的情绪状态，提供更加全面的看护服务。The introduction of emotion recognition function can provide a more comprehensive and in-depth understanding and attention to the health status of the elderly. In addition to physical health and behavioral patterns, it is now possible to pay attention to the emotional state of the elderly and provide more comprehensive care services.
情绪状态的异常变化可能是健康问题的前兆。通过实时监测和识别老人的情绪变化，可以提前发现可能的健康风险，及时进行干预。Abnormal changes in emotional state may be a precursor to health problems. By real-time monitoring and identifying the emotional changes of the elderly, possible health risks can be discovered in advance and timely intervention can be made.
通过个性化的音频播放，根据老人的情绪状态播放适当的声音或音乐，可以有效地改善老人的情绪，提高他们的生活质量。Through personalized audio playback, playing appropriate sounds or music according to the emotional state of the elderly can effectively improve the elderly's mood and enhance their quality of life.
声音的音色可以根据老人家人的声音进行特别设置，这种熟悉的声音更有可能引起老人的积极反应，增强了系统的人性化设计。The timbre of the sound can be specially set according to the voices of the elderly’s family members. This familiar sound is more likely to cause a positive response from the elderly, enhancing the humanized design of the system.
利用卷积神经网络(CNN)和长短期记忆网络(LSTM)的结合，可以有效地处理复杂的语音数据，精准地识别出老人的情绪状态。The combination of convolutional neural network (CNN) and long short-term memory network (LSTM) can effectively process complex speech data and accurately identify the emotional state of the elderly.
通过数据处理与控制模块结合预警预测信息，系统可以启动辅助设备，如智能机器人或智能座椅，对风险较高的老人进行重点看护。这种智能看护不仅仅局限于生理和行为，更加关注老人的心理健康和情绪状况，从而提升了系统的智能程度。By combining the data processing and control module with the early warning and prediction information, the system can activate auxiliary equipment, such as intelligent robots or intelligent seats, to provide key care for the elderly at higher risk. This intelligent care is not limited to physiology and behavior, but pays more attention to the mental health and emotional state of the elderly, thereby improving the intelligence of the system.
附图说明BRIEF DESCRIPTION OF THE DRAWINGS
图1是本发明系统示意图。FIG. 1 is a schematic diagram of the system of the present invention.
具体实施方式DETAILED DESCRIPTION
下面结合附图对本发明的具体实施方式作描述。The specific implementation of the present invention will be described below in conjunction with the accompanying drawings.
如图1所示，本发明系统包括如下组成部分：As shown in Figure 1, the system of the present invention includes the following components:
传感器模块：这个模块包括高精度毫米波雷达传感器、语音传感器以及老人随身携带的生物传感器。这些传感器可以在老人的生活区域内安装。毫米波雷达传感器和语音传感器可以通过WiFi连接到数据处理与控制模块，而生物传感器可以通过蓝牙或其他近场通信技术连接到数据处理与控制模块。Sensor module: This module includes high-precision millimeter-wave radar sensors, voice sensors, and biosensors that the elderly can carry with them. These sensors can be installed in the elderly's living area. The millimeter-wave radar sensor and voice sensor can be connected to the data processing and control module via WiFi, while the biosensor can be connected to the data processing and control module via Bluetooth or other near-field communication technologies.
数据处理与控制模块：这个模块负责接收和处理传感器模块提供的数据，并用于生成预警预测信息。数据处理与控制模块可以运行在一个服务器上，服务器可以位于云端或用户的家庭网络中。云端部署的优点是能够利用大数据计算能力，提供稳定的服务，同时易于数据的集中管理和分析。家庭网络中部署的优点是数据的传输延迟低，且数据在本地处理，有利于保护用户隐私。Data processing and control module: This module is responsible for receiving and processing the data provided by the sensor module and is used to generate early warning and prediction information. The data processing and control module can run on a server, which can be located in the cloud or in the user's home network. The advantage of cloud deployment is that it can utilize big data computing capabilities to provide stable services, while facilitating centralized management and analysis of data. The advantage of deployment in the home network is that the data transmission delay is low and the data is processed locally, which is conducive to protecting user privacy.
报警模块：这个模块通过内置的无线通信装置与老人家属或救护中心进行通信。这个模块可能需要有一个内置的SIM卡，能够通过2G/3G/4G/5G网络进行通信。Alarm module: This module communicates with the elderly’s family or rescue center through a built-in wireless communication device. This module may need to have a built-in SIM card and be able to communicate through 2G/3G/4G/5G networks.
交流模块：这个模块包括设置在老人生活区域的音箱或老人随身佩戴的语音装置。音箱可以通过WiFi连接到数据处理与控制模块，而随身佩戴的语音装置可以通过蓝牙或其他近场通信技术连接到数据处理与控制模块。Communication module: This module includes a speaker set up in the elderly's living area or a voice device worn by the elderly. The speaker can be connected to the data processing and control module via WiFi, while the voice device worn by the elderly can be connected to the data processing and control module via Bluetooth or other near-field communication technologies.
首先说明情绪识别模型，包括：First, the emotion recognition model is described, including:
语音收集设备(传感器模块内包括)：这部分可以由在老人生活区域放置的麦克风或者老人随身携带的语音收集装置来实现。这些设备应该具有高质量的音频采集能力，并且能够在各种环境条件下稳定工作。Voice collection equipment (included in the sensor module): This part can be achieved by microphones placed in the elderly’s living area or voice collection devices carried by the elderly. These devices should have high-quality audio collection capabilities and be able to work stably under various environmental conditions.
语音预处理模块：这部分主要负责对收集到的语音进行预处理。预处理可能包括降噪、分段、语音增强等步骤，以便将语音信号转化为更适合后续处理的形式。具体操作可以采用如Python中librosa库等现有的音频处理工具。Speech preprocessing module: This part is mainly responsible for preprocessing the collected speech. Preprocessing may include steps such as noise reduction, segmentation, and speech enhancement to convert the speech signal into a form more suitable for subsequent processing. The specific operation can be carried out using existing audio processing tools such as the librosa library in Python.
特征提取模块：这部分主要负责从预处理后的语音中提取特征。这里我们采用Mel频率倒谱系数(MFCC)作为特征，这是在语音和语音识别中广泛使用的特征。特征提取可以用到例如Python的librosa库等工具。Feature extraction module: This part is mainly responsible for extracting features from the preprocessed speech. Here we use Mel Frequency Cepstral Coefficient (MFCC) as features, which are widely used in speech and voice recognition. Feature extraction can be done using tools such as Python's librosa library.
情绪分类模型：这部分是系统的核心，负责将提取出来的特征输入到情绪分类模型中，然后输出情绪类别。模型采用卷积神经网络(CNN)和长短期记忆网络(LSTM)的结合来进行处理。具体设计可以是：Emotion classification model: This part is the core of the system, responsible for inputting the extracted features into the emotion classification model and then outputting the emotion category. The model uses a combination of convolutional neural network (CNN) and long short-term memory network (LSTM) for processing. The specific design can be:
输入层接受MFCC特征的二维数组，其大小可以是[时间步长,MFCC特征数]，例如[100,13]。The input layer accepts a 2D array of MFCC features, whose size can be [timestep, number of MFCC features], for example [100, 13].
接着是一维的卷积层(例如有128个卷积核，卷积核大小为5)和最大池化层，用于提取语音的局部频率模式。This is followed by a one-dimensional convolutional layer (e.g., with 128 convolution kernels and a kernel size of 5) and a maximum pooling layer to extract the local frequency pattern of the speech.
然后是一个LSTM层(例如有64个单元)，用于捕获长期的时间依赖性。This is followed by an LSTM layer (e.g. with 64 units) to capture long-term temporal dependencies.
最后是一个全连接层(例如有5个神经元，代表5个情绪类别)，用于情绪分类。输出层应用softmax激活函数，使得输出可以解释为各类别的概率。Finally, there is a fully connected layer (for example, 5 neurons, representing 5 emotion categories) for emotion classification. The output layer applies a softmax activation function so that the output can be interpreted as the probability of each category.
训练和验证：为了让我们的模型能正确地进行情绪分类，我们需要有标签的数据进行训练。我们可以收集一些老人的语音样本，然后由专业人员进行情绪标注。训练过程中，我们通常会设置一部分数据作为验证集，用来评估模型的性能和调整超参数。具体来说：Training and validation: In order for our model to correctly classify emotions, we need labeled data for training. We can collect some speech samples from the elderly and have professionals label the emotions. During the training process, we usually set a portion of the data as a validation set to evaluate the performance of the model and adjust the hyperparameters. Specifically:
在训练模型之前，我们需要预处理数据。对于音频数据，这可能包括噪声削减、标准化、分帧、特征提取等步骤。Before training the model, we need to preprocess the data. For audio data, this may include steps such as noise reduction, normalization, framing, feature extraction, etc.
特征提取：提取音频数据的特征，如音高、语速、音色、韵律等，作为模型的输入。Feature extraction: Extract the features of audio data, such as pitch, speaking speed, timbre, rhythm, etc., as input to the model.
模型选择和训练：选择合适的机器学习或深度学习模型进行训练。常见的模型包括SVM(支持向量机)、随机森林、神经网络等。训练的目标是使模型学习到从特征到情绪类别的映射关系。这个过程可能需要大量的计算资源和时间。Model selection and training: Select an appropriate machine learning or deep learning model for training. Common models include SVM (support vector machine), random forest, neural network, etc. The goal of training is to enable the model to learn the mapping relationship from features to emotion categories. This process may require a lot of computing resources and time.
模型验证：在验证集上验证模型的性能。这可以通过计算模型的精度、召回率、F1值等指标来完成。如果性能不满意，可能需要回到模型选择和训练的步骤，选择不同的模型或调整模型的参数。Model Validation: Validate the performance of the model on the validation set. This can be done by calculating the model's precision, recall, F1 value, and other metrics. If the performance is not satisfactory, you may need to go back to the model selection and training steps and choose a different model or adjust the model's parameters.
模型调优：在训练过程中，我们需要监控模型的学习曲线，以便在模型过拟合或者欠拟合时及时调整。可能需要调整的参数包括学习率、批次大小、优化器类型等。Model tuning: During the training process, we need to monitor the learning curve of the model so that we can make timely adjustments when the model is overfitting or underfitting. Parameters that may need to be adjusted include learning rate, batch size, optimizer type, etc.
交叉验证：为了确保模型的泛化性能，可以使用交叉验证的方式进行模型训练和验证。比如k-折交叉验证，将数据集分为k个部分，每次取一个部分作为验证集，其他部分作为训练集，总共进行k次训练和验证。Cross-validation: To ensure the generalization performance of the model, you can use cross-validation to train and validate the model. For example, k-fold cross-validation divides the data set into k parts, takes one part as the validation set each time, and the other parts as the training set, and performs k training and validation times in total.
测试：使用未参与训练和验证的测试集，对模型进行最终的性能评估。Testing: Use the test set that was not used in training and validation to perform a final performance evaluation of the model.
训练与验证结束后，还需要：After training and validation, you also need to:
后处理模块：这部分负责将模型的输出转化为用户可以理解的形式。如果输出是一个情绪类别的概率分布，后处理模块可能需要选择概率最大的类别作为预测的情绪。此外，后处理模块还可以进行其他处理，如情绪的平滑化(例如，基于时间序列的情绪变化模式，避免过于频繁的情绪转换)。Post-processing module: This part is responsible for converting the output of the model into a form that the user can understand. If the output is a probability distribution of emotion categories, the post-processing module may need to select the category with the highest probability as the predicted emotion. In addition, the post-processing module can also perform other processing, such as smoothing emotions (for example, based on the emotion change pattern of the time series, to avoid too frequent emotion transitions).
用户界面(UI)：这部分主要负责将系统的输出显示给用户，并收集用户的反馈。界面设计应考虑用户友好，易读易懂。可以显示历史情绪变化趋势图、实时情绪等信息。User Interface (UI): This part is mainly responsible for displaying the system output to the user and collecting user feedback. The interface design should be user-friendly and easy to read and understand. It can display historical sentiment trend charts, real-time sentiment and other information.
反馈系统：这部分负责处理用户的反馈。反馈可能包括对系统输出的评价、对系统操作的建议等。反馈系统可以帮助我们改进模型，提高系统的准确性和用户满意度。Feedback system: This part is responsible for processing user feedback. Feedback may include evaluation of system output, suggestions for system operation, etc. The feedback system can help us improve the model, improve system accuracy and user satisfaction.
在另一个实施例中，本发明还包括以下内容:In another embodiment, the present invention further includes the following:
数据处理与控制模块利用医疗大数据和历史医疗数据，结合实时生理数据和情绪数据的变化，预测老人的行为并生成预警预测信息；数据处理与控制模块连接报警模块和提醒模块；其具体可以包括如下一种实施方式：The data processing and control module uses medical big data and historical medical data, combined with changes in real-time physiological data and emotional data, to predict the behavior of the elderly and generate early warning prediction information; the data processing and control module is connected to the alarm module and the reminder module; it may specifically include the following implementation methods:
1.数据预处理：其中包括标准化、归一化、去噪、填充缺失值等步骤。这个步骤很关键，因为它直接影响到模型的训练效果。预处理模块会用于处理来自传感器模块的实时生理数据和情绪数据，以及医疗大数据和历史医疗数据。1. Data preprocessing: This includes standardization, normalization, denoising, and filling missing values. This step is critical because it directly affects the training effect of the model. The preprocessing module is used to process real-time physiological data and emotional data from the sensor module, as well as medical big data and historical medical data.
2.特征提取：从预处理后的数据中提取有用的特征，这些特征将被用于训练模型。这可能包括统计特征(如平均值、中位数、方差等)、频域特征(如傅里叶变换或小波变换的系数)等。2. Feature extraction: Extract useful features from the preprocessed data, which will be used to train the model. This may include statistical features (such as mean, median, variance, etc.), frequency domain features (such as Fourier transform or wavelet transform coefficients), etc.
3.训练预测模型：一旦获得了特征，就可以开始训练预测模型。这可能包括监督学习算法(如决策树、支持向量机、随机森林、神经网络等)和无监督学习算法(如聚类、自编码器等)。模型的选择将取决于问题的性质和数据的特征。3. Training the predictive model: Once the features are obtained, you can start training the predictive model. This may include supervised learning algorithms (such as decision trees, support vector machines, random forests, neural networks, etc.) and unsupervised learning algorithms (such as clustering, autoencoders, etc.). The choice of model will depend on the nature of the problem and the characteristics of the data.
4.行为预测：一旦训练了预测模型，就可以使用它来预测老人的行为。这将涉及模型的推理，即使用模型处理新的数据并生成预测结果。4. Behavior prediction: Once the prediction model is trained, it can be used to predict the behavior of the elderly. This will involve reasoning about the model, that is, using the model to process new data and generate prediction results.
5.预警预测信息生成：基于预测的行为，数据处理与控制模块会生成预警预测信息。如果预测的行为超出了某个阈值或表现出某些特定的模式，那么预警信息将被生成。5. Generation of early warning prediction information: Based on the predicted behavior, the data processing and control module will generate early warning prediction information. If the predicted behavior exceeds a certain threshold or shows certain specific patterns, then early warning information will be generated.
6.连接报警模块和提醒模块：数据处理与控制模块需要与报警模块和提醒模块连接。当预警信息生成时，这个模块会触发报警模块发送警报，或者触发提醒模块给老人或其家属发送提醒。6. Connect the alarm module and reminder module: The data processing and control module needs to be connected to the alarm module and reminder module. When the warning information is generated, this module will trigger the alarm module to send an alarm, or trigger the reminder module to send a reminder to the elderly or their family members.
在另一个实施例中，本发明还包括数据处理与控制模块结合预警预测信息，启动辅助设备对风险老人进行重点看护，包括情感对话、提醒、辅助机器人或智能座椅。In another embodiment, the present invention further includes a data processing and control module that combines the early warning prediction information to activate auxiliary equipment to provide key care for the elderly at risk, including emotional dialogue, reminders, auxiliary robots or smart seats.
其中，情感对话、提醒如上所述，辅助机器人应该具有多种功能以满足老年人的需求，例如：Among them, emotional dialogue, reminder As mentioned above, the assistive robot should have multiple functions to meet the needs of the elderly, such as:
移动功能：机器人应该有能力在家庭环境中自由移动，能够避开障碍物，并且能够在需要的时候帮助老人移动。这可以通过使用高级导航和路径规划算法实现。Mobility: The robot should have the ability to move freely in the home environment, avoid obstacles, and be able to assist the elderly in moving when needed. This can be achieved by using advanced navigation and path planning algorithms.
物品搬运功能：机器人需要能够搬运物品，如食物、药品等。这可能需要柔性抓取器以安全地搬运各种形状和大小的物品。Object handling capabilities: The robot needs to be able to carry objects such as food, medicine, etc. This may require flexible grippers to safely carry objects of various shapes and sizes.
语音识别和反馈功能：机器人应该能够理解并回应语音指令，这可能需要集成高级的语音识别和自然语言理解系统。Speech recognition and feedback capabilities: The robot should be able to understand and respond to voice commands, which may require the integration of advanced speech recognition and natural language understanding systems.
监控和紧急响应功能：机器人应该能够监控老人的生理状况和行为，如果检测到可能的紧急情况，如跌倒或者其他医疗紧急情况，应该能够迅速作出反应。Monitoring and emergency response capabilities: The robot should be able to monitor the elderly’s physiological condition and behavior and respond quickly if it detects a possible emergency, such as a fall or other medical emergency.
智能座椅应该具有一些关键特性，如：Smart seats should have some key features, such as:
舒适度调整：座椅应该能够根据老人的身体形状和坐姿自动调整，保持最高的舒适度。这可能需要使用压力传感器来检测老人的坐姿，并使用电动驱动器来改变座椅的形状。Comfort adjustment: The seat should be able to automatically adjust to the elderly's body shape and sitting posture to maintain the highest comfort. This may require the use of pressure sensors to detect the elderly's sitting posture and electric actuators to change the shape of the seat.
生理参数监测：座椅可以集成传感器来实时监测老人的生理参数，如心率、血压、体温等。如果这些参数超出正常范围，座椅可以向数据处理模块发送警告。Physiological parameter monitoring: The seat can be integrated with sensors to monitor the elderly's physiological parameters in real time, such as heart rate, blood pressure, body temperature, etc. If these parameters are out of the normal range, the seat can send a warning to the data processing module.
活动提醒：如果老人长时间保持同一坐姿，座椅可以发出提醒，鼓励老人起身活动或改变坐姿，以防止身体僵硬或压疮。Activity reminder: If the elderly maintain the same sitting position for a long time, the chair can issue a reminder to encourage them to get up and move around or change their sitting position to prevent body stiffness or pressure sores.
紧急情况响应：如果检测到可能的紧急情况，如心率异常，座椅应该能够迅速作出反应，比如立即向数据处理模块发送警告，或者使用内置的通讯设备呼叫紧急联系人。Emergency response: If a possible emergency situation is detected, such as an abnormal heart rate, the seat should be able to respond quickly, such as immediately sending a warning to the data processing module or calling an emergency contact using the built-in communication device.
这些设计都需要集成到数据处理与控制模块中，以确保它们能够根据预测的信息有效地作出响应。These designs need to be integrated into data processing and control modules to ensure that they can respond effectively based on the predicted information.
除此之外，在另一些实施例中，本发明还可以包括如下步骤：In addition, in some other embodiments, the present invention may further include the following steps:
S1：传感器模块将收集到的有关老人的传感器数据及对应的时间发送到数据处理与控制模块；S1: The sensor module sends the collected sensor data about the elderly and the corresponding time to the data processing and control module;
S2：每当数据处理与控制模块识别到传感器数据的变化速率超过预设阈值时，通过交流模块询问老人已完成的行为以及其该行为的健康程度，并结合询问结果以及已接收到的传感器数据，获取传感器数据与老人行为的关联关系，并同时标记该行为的发生时间段以及对应的健康程度；当数据处理与控制模块的所接收到的数据量积累到预设数据量时，进入步骤S3～S5；S2: Whenever the data processing and control module identifies that the rate of change of the sensor data exceeds a preset threshold, the communication module asks the elderly about the completed behavior and the health level of the behavior, and combines the inquiry result with the received sensor data to obtain the correlation between the sensor data and the elderly's behavior, and simultaneously marks the time period of the behavior and the corresponding health level; when the amount of data received by the data processing and control module accumulates to the preset amount of data, it enters steps S3 to S5;
S3：数据处理与控制模块对老人的所有行为及其对应的发生时间进行关联得到一系列行为序列作为老人的总体行为模式；S3: The data processing and control module associates all behaviors of the elderly and their corresponding occurrence times to obtain a series of behavior sequences as the overall behavior pattern of the elderly;
S4：数据处理与控制模块对老人的超过健康阈值的所有行为及其对应的发生时间进行关联得到一系列行为序列作为老人的健康行为模式；S4: The data processing and control module associates all behaviors of the elderly that exceed the health threshold and their corresponding occurrence time to obtain a series of behavior sequences as the healthy behavior pattern of the elderly;
S5：数据处理与控制模块对老人的低于健康阈值的所有行为及其对应的发生时间进行关联得到一系列行为序列作为老人的非健康行为模式；S5: The data processing and control module associates all behaviors of the elderly below the health threshold and their corresponding occurrence times to obtain a series of behavior sequences as the elderly's unhealthy behavior patterns;
S6：当数据处理与控制模块得到老人的总体行为模式、健康行为模式和非健康行为模式时，通过交流模块提示老人可进入看护阶段；S6: When the data processing and control module obtains the overall behavior pattern, healthy behavior pattern and unhealthy behavior pattern of the elderly, the communication module prompts the elderly to enter the care stage;
S7：在看护阶段中，数据处理与控制模块实时获取传感器数据，并结合S2中的关联关系识别传感器数据对应的行为，并取预先设置的滑动窗口时间段内的随时间变化的行为序列与总体行为模式、健康行为模式以及非健康行为模式进行匹配，获取三个相似度，并将三个相似度中最高的相似度与预设相似度阈值进行比对；S7: In the care phase, the data processing and control module acquires sensor data in real time, identifies the behavior corresponding to the sensor data in combination with the association relationship in S2, and matches the behavior sequence that changes over time within a preset sliding window period with the overall behavior pattern, the healthy behavior pattern, and the unhealthy behavior pattern, obtains three similarities, and compares the highest similarity among the three similarities with the preset similarity threshold;
若三个相似度中，对应于健康行为模式的相似度最高且超过预设的匹配阈值，则数据处理与控制模块通过交流模块鼓励老人；If, among the three similarities, the similarity corresponding to the healthy behavior pattern is the highest and exceeds the preset matching threshold, the data processing and control module encourages the elderly through the communication module;
若三个相似度中，对应于非健康行为模式的相似度最高且超过预设的匹配阈值，则数据处理与控制模块通过交流模块提示老人要注意行为健康、播报老人在该滑动窗口内的所有行为，并依据健康行为模式对老人行为的预测给予老人实时提醒、生活运动建议和饮食建议；If the similarity corresponding to the unhealthy behavior pattern is the highest among the three similarities and exceeds the preset matching threshold, the data processing and control module will remind the elderly to pay attention to behavioral health through the communication module, broadcast all the elderly's behaviors within the sliding window, and give the elderly real-time reminders, life exercise suggestions and diet suggestions based on the prediction of the healthy behavior pattern;
若三个相似度均不超过预设相似度阈值，则通过提示模块询问老人在滑动窗口内进行了哪些行为，若未得到老人的回应，则数据处理与控制模块通过报警模块与老人家属或救护中心进行报警；若得到老人回应，则数据处理与控制模块将该滑动窗口内的传感器数据及对应的行为、行为发生时间加入至步骤S2～S5中所采用的训练数据，进而采用增量学习的方式实时调整S2中的关联关系以及S3～S5中的三种行为模式，并将其实时应用至老人的看护阶段。If the three similarities do not exceed the preset similarity threshold, the prompt module will ask the elderly what behaviors they have performed in the sliding window. If there is no response from the elderly, the data processing and control module will alert the elderly's family or the rescue center through the alarm module; if the elderly responds, the data processing and control module will add the sensor data and the corresponding behaviors and the time when the behaviors occurred in the sliding window to the training data used in steps S2 to S5, and then use incremental learning to adjust the association relationship in S2 and the three behavior patterns in S3 to S5 in real time, and apply them in real time to the care stage of the elderly.
健康程度通过健康程度因子评估，健康程度因子取0～100，数值越高表示行为越健康。Health level is assessed through the health level factor, which ranges from 0 to 100, with higher values indicating healthier behavior.
在S7步骤中，“依据健康行为模式对老人行为的预测”还可以通过以下方式来实现：In step S7, "prediction of elderly behavior based on health behavior patterns" can also be achieved in the following ways:
我们可以根据老人过往的健康行为模式建立一个预测模型。例如，如果我们观察到当老人每日早晨进行轻度锻炼(如慢跑或散步)时，他们的生理参数(如心率、血压等)保持在一个健康的范围内，我们可以将这种行为视为健康行为模式。同样，我们也可以观察到其他的健康行为模式，例如，保持良好的饮食习惯、保持充足的睡眠时间等。当然也可以结合上述的询问老人的方式获取健康情况。We can build a prediction model based on the elderly's past health behavior patterns. For example, if we observe that when the elderly do light exercise (such as jogging or walking) every morning, their physiological parameters (such as heart rate, blood pressure, etc.) remain within a healthy range, we can regard this behavior as a healthy behavior pattern. Similarly, we can also observe other healthy behavior patterns, such as maintaining good eating habits, getting enough sleep, etc. Of course, we can also combine the above-mentioned way of asking the elderly to obtain health conditions.
建立预测模型后，我们可以实时监测老人的行为数据，与我们的预测模型进行比较。如果我们观察到老人的当前行为与健康行为模式不符，例如，减少了锻炼时间，或者开始摄入高脂肪食物，我们可以提醒他们注意，给予相应的建议，以鼓励他们回归到健康行为模式。同时，数据处理与控制模块也会实时调整预测模型，以更好地适应老人的行为变化。After establishing the prediction model, we can monitor the elderly's behavior data in real time and compare it with our prediction model. If we observe that the elderly's current behavior is inconsistent with the healthy behavior pattern, for example, they have reduced their exercise time or started to eat high-fat foods, we can remind them to pay attention and give corresponding suggestions to encourage them to return to healthy behavior patterns. At the same time, the data processing and control module will also adjust the prediction model in real time to better adapt to the elderly's behavioral changes.
这个过程可以通过机器学习的方法实现。我们可以使用老人的历史行为数据和生理参数数据来训练一个预测模型，例如，可以使用支持向量机(SVM)、随机森林(RandomForest)、深度学习(Deep Learning)等方法。训练好的模型可以对老人未来的行为进行预测，当预测到有潜在的健康风险时，可以提前进行预警，从而进行早期干预，防止健康问题的发生。This process can be achieved through machine learning. We can use the historical behavior data and physiological parameter data of the elderly to train a prediction model, for example, we can use support vector machine (SVM), random forest (RandomForest), deep learning (Deep Learning) and other methods. The trained model can predict the future behavior of the elderly. When potential health risks are predicted, early warning can be given, so as to carry out early intervention and prevent the occurrence of health problems.
以下表格表示在一些实施例中测得的一些数据：The following table shows some data measured in some embodiments:
以下表格1中，列出了在智能看护系统中使用的主要传感器，包括毫米波雷达传感器、语音传感器和生物传感器，它们的主要功能是监测老人的生理参数和行为状态，如行动模式、语音模式等：The following Table 1 lists the main sensors used in the smart care system, including millimeter wave radar sensors, voice sensors and biosensors. Their main function is to monitor the physiological parameters and behavioral status of the elderly, such as action patterns, voice patterns, etc.:
以下表格2描述了数据处理和控制模块的输入和输出。输入包括从传感器模块获取的实时生理参数和情绪数据，以及历史医疗数据。输出则是基于这些数据生成的预警预测信息：Table 2 below describes the input and output of the data processing and control module. The input includes real-time physiological parameters and emotional data obtained from the sensor module, as well as historical medical data. The output is the early warning prediction information generated based on these data:
以下表格3展示了报警模块的主要功能，其输入是来自数据处理和控制模块的预警预测信息，当发现问题时，报警模块将通过无线通信装置与老人家属或救护中心进行通信：Table 3 below shows the main functions of the alarm module. Its input is the early warning prediction information from the data processing and control module. When a problem is found, the alarm module will communicate with the elderly’s family or the rescue center through a wireless communication device:
以下表格4列出了交流模块的主要功能。当数据处理与控制模块识别老人的情绪状态后，交流模块会根据老人的情绪状态播放特定的音乐或者声音，目的是帮助改善或稳定老人的情绪状态：The following table 4 lists the main functions of the communication module. After the data processing and control module recognizes the emotional state of the elderly, the communication module will play specific music or sounds according to the emotional state of the elderly, in order to help improve or stabilize the emotional state of the elderly:
以上所有表格中，“N/A”是“Not Applicable”的缩写，意味着“不适用”。在表格中，N/A通常用于指示某个特定字段对于给定的情况或者条件不适用或者没有数据可供提供。也就是说，这个单元格并不需要填写信息，或者信息暂时无法获取。In all the above tables, "N/A" is the abbreviation of "Not Applicable", which means "not applicable". In tables, N/A is usually used to indicate that a specific field is not applicable to a given situation or condition or that no data is available. In other words, this cell does not need to be filled in with information, or the information is temporarily unavailable.
以上所述，仅为本发明较佳的具体实施方式，但本发明的保护范围并不局限于此，任何熟悉本技术领域的技术人员在本发明披露的技术范围内，根据本发明的技术方案及其发明构思加以等同替换或改变，都应涵盖在本发明的保护范围之内。The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310637336.8ACN116646069B (en) | 2023-06-01 | 2023-06-01 | Intelligent care system and care method for aged people based on internet of things (IoT) |
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310637336.8ACN116646069B (en) | 2023-06-01 | 2023-06-01 | Intelligent care system and care method for aged people based on internet of things (IoT) |
| Publication Number | Publication Date |
|---|---|
| CN116646069Atrue CN116646069A (en) | 2023-08-25 |
| CN116646069B CN116646069B (en) | 2023-12-19 |
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310637336.8AActiveCN116646069B (en) | 2023-06-01 | 2023-06-01 | Intelligent care system and care method for aged people based on internet of things (IoT) |
| Country | Link |
|---|---|
| CN (1) | CN116646069B (en) |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117695103A (en)* | 2023-11-20 | 2024-03-15 | 电子科技大学 | A multi-point pressure detection and pressure ulcer early warning system based on flexible pressure sensors |
| CN118113198A (en)* | 2024-04-29 | 2024-05-31 | 深圳市爱保护科技有限公司 | AI dial control method and system based on intelligent bracelet |
| CN118312857A (en)* | 2024-06-04 | 2024-07-09 | 广东海洋大学 | A multimodal emotion recognition method and system |
| CN120431677A (en)* | 2025-07-08 | 2025-08-05 | 运城市恩光科技有限公司 | Collaborative emergency response method, system and equipment based on smart wearable devices |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103325080A (en)* | 2013-06-21 | 2013-09-25 | 电子科技大学 | Gerocamium intelligent nursing system and method based on Internet of Things technology |
| CN105404777A (en)* | 2015-11-14 | 2016-03-16 | 合肥骇虫信息科技有限公司 | Multi-functional smart home care system |
| CN109326081A (en)* | 2018-11-09 | 2019-02-12 | 复旦大学 | IoT-based home care early warning system and health status assessment method for the elderly |
| CN111079440A (en)* | 2019-12-12 | 2020-04-28 | 东南大学 | Old man attends to robot subsystem based on emotion recognition |
| US20220084543A1 (en)* | 2020-01-21 | 2022-03-17 | Rishi Amit Sinha | Cognitive Assistant for Real-Time Emotion Detection from Human Speech |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103325080A (en)* | 2013-06-21 | 2013-09-25 | 电子科技大学 | Gerocamium intelligent nursing system and method based on Internet of Things technology |
| CN105404777A (en)* | 2015-11-14 | 2016-03-16 | 合肥骇虫信息科技有限公司 | Multi-functional smart home care system |
| CN109326081A (en)* | 2018-11-09 | 2019-02-12 | 复旦大学 | IoT-based home care early warning system and health status assessment method for the elderly |
| CN111079440A (en)* | 2019-12-12 | 2020-04-28 | 东南大学 | Old man attends to robot subsystem based on emotion recognition |
| US20220084543A1 (en)* | 2020-01-21 | 2022-03-17 | Rishi Amit Sinha | Cognitive Assistant for Real-Time Emotion Detection from Human Speech |
| Title |
|---|
| 张家铭;王晓曼;景文博;: "基于深度卷积网络和谱图的语音情感识别", 长春理工大学学报(自然科学版), no. 01* |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117695103A (en)* | 2023-11-20 | 2024-03-15 | 电子科技大学 | A multi-point pressure detection and pressure ulcer early warning system based on flexible pressure sensors |
| CN118113198A (en)* | 2024-04-29 | 2024-05-31 | 深圳市爱保护科技有限公司 | AI dial control method and system based on intelligent bracelet |
| CN118312857A (en)* | 2024-06-04 | 2024-07-09 | 广东海洋大学 | A multimodal emotion recognition method and system |
| CN120431677A (en)* | 2025-07-08 | 2025-08-05 | 运城市恩光科技有限公司 | Collaborative emergency response method, system and equipment based on smart wearable devices |
| Publication number | Publication date |
|---|---|
| CN116646069B (en) | 2023-12-19 |
| Publication | Publication Date | Title |
|---|---|---|
| CN116646069B (en) | Intelligent care system and care method for aged people based on internet of things (IoT) | |
| CN108197115A (en) | Intelligent interactive method, device, computer equipment and computer readable storage medium | |
| Chang et al. | An Efficient Classification of Neonates Cry Using Extreme Gradient Boosting‐Assisted Grouped‐Support‐Vector Network | |
| AU2021206060A1 (en) | Dynamic user response data collection method | |
| CN116564561A (en) | Intelligent voice nursing system and nursing method based on physiological and emotion characteristics | |
| Alshammari | Applying machine learning algorithms for the classification of sleep disorders | |
| JPWO2018168369A1 (en) | Machine learning device and machine learning program | |
| CN112135568B (en) | Information processing device, information processing method, and program | |
| WO2014042599A1 (en) | System and method for developing a model indicative of a subject's emotional state when listening to musical pieces | |
| CN118427622A (en) | Health care training method and system based on virtual simulation technology | |
| Abrar et al. | A multi-agent approach for personalized hypertension risk prediction | |
| CN118787831A (en) | Emotion perception and intervention system, method, device and robot | |
| Rezaee et al. | Can you understand why i am crying? a decision-making system for classifying infants’ cry languages based on deepsvm model | |
| CN118645264A (en) | A non-contact, non-invasive and non-destructive sugar monitoring, early warning and intervention method | |
| CN118692662A (en) | An audio and video recognition system based on multimodal analysis | |
| CN117668648A (en) | A pet status recognition device based on artificial intelligence technology | |
| Özseven | A review of infant cry recognition and classification based on computer-aided diagnoses | |
| Talal et al. | Machine learning‐based classification of multiple heart disorders from PCG signals | |
| Pandey et al. | Nocturnal sleep sounds classification with artificial neural network for sleep monitoring | |
| Vaishnavi et al. | Premature Infant Cry Classification via Elephant Herding Optimized Convolutional Gated Recurrent Neural Network | |
| Mala et al. | Effective infant cry signal analysis and reasoning using IARO based leaky Bi-LSTM model | |
| Sanjeev et al. | Deep learning-based mixed data approach for COVID-19 detection | |
| Sulthana et al. | Parkinson's Disease Prediction using XGBoost and SVM | |
| Gulsen et al. | Machine Learning Models With Hyperparameter Optimization for Voice Pathology Classification on Saarbrücken Voice Database | |
| Jahin et al. | A modern approach to AI assistant for heart disease detection by heart sound through created e-Stethoscope |
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |