CN111563663A - Robot, service quality evaluation method and system - Google Patents

Abstract

The application discloses a robot, a service quality evaluation method and a service quality evaluation system. The robot provided by the embodiment of the application is provided with a binocular camera, an action analysis module, a voice emotion analysis module and an image emotion analysis module, and the satisfaction degree of a client and the service level of a service worker can be obtained according to a trained deep convolutional neural network and serve as first grading data; obtaining the degree of environment tidiness as second grading data according to the environment information; generating third grading data according to the response speed of the service personnel in the non-service area; according to the action data of the client in the non-service area, the satisfaction degree of the client on the work efficiency of service personnel can be obtained, fourth scoring data is generated, the service quality evaluation is comprehensively obtained through the multi-type data, the GPS module is arranged, the obstacle avoidance route can be obtained, the service quality evaluation can be obtained by the robot under the condition that the user is not affected, and the reference value of the service quality evaluation is effectively improved.

Description

Translated fromChinese

一种机器人、服务质量评价方法及系统A kind of robot, service quality evaluation method and system

技术领域technical field

本申请涉及数据处理技术领域，特别是一种机器人、服务质量评 价方法及系统。The present application relates to the technical field of data processing, in particular to a robot, a service quality evaluation method and system.

背景技术Background technique

对于服务机构而言，服务人员的素质和服务环境是服务质量的重 要保证。传统的服务质量评价方法主要依靠客户打分或者问卷调查， 然而客户打分通常比较随意，问卷调查耗费的人力物力较多，效率也 不高。为了实现服务质量评价的自动化，市面上出现了一些服务质量 评价机器人，能够通过面部表情识别出客户的情绪，从而进行自动评 价，但是对服务质量的影响不仅包括服务过程，现有技术只能对服务 人员的服务过程进行自动评价，不能代表整个服务机构的服务质量， 参考价值不高。For service organizations, the quality of service personnel and service environment are important guarantees for service quality. Traditional service quality evaluation methods mainly rely on customer scores or questionnaires. However, customer scores are usually more arbitrary, and questionnaire surveys consume a lot of manpower and material resources, and the efficiency is not high. In order to realize the automation of service quality evaluation, some service quality evaluation robots have appeared on the market, which can identify customers' emotions through facial expressions, so as to carry out automatic evaluation, but the impact on service quality not only includes the service process, the existing technology can only The automatic evaluation of the service process of the service personnel cannot represent the service quality of the entire service organization, and the reference value is not high.

发明内容SUMMARY OF THE INVENTION

为了克服现有技术的不足，本申请的目的在于提供一种机器人、 服务质量评价方法及系统，能够从自动获取多种类型的服务质量评价 数据，提高服务质量评价的全面性。In order to overcome the deficiencies of the prior art, the purpose of this application is to provide a robot, a service quality evaluation method and system, which can automatically obtain various types of service quality evaluation data and improve the comprehensiveness of service quality evaluation.

本申请解决其问题所采用的技术方案是：第一方面，本申请提供 了一种机器人，所述机器人包括：The technical solution adopted by the application to solve its problems is: in the first aspect, the application provides a robot, and the robot includes:

双目摄像头，用于获取面部图像信息、动作信息、环境信息、客 户位置信息、距离信息；Binocular camera, used to obtain facial image information, motion information, environmental information, customer location information, distance information;

语音接收模块，用于接收语音信息；A voice receiving module for receiving voice information;

动作分析模块，用于根据所述动作信息和所获取的深度卷积神经 网络获取动作数据；an action analysis module for acquiring action data according to the action information and the acquired deep convolutional neural network;

语音情绪分析模块，用于根据所述语音信息和所获取的所述深度 卷积神经网络获取语音情绪数据；A speech emotion analysis module for acquiring speech emotion data according to the speech information and the acquired deep convolutional neural network;

图像情绪分析模块，用于根据所述面部图像信息和所获取的所述 深度卷积神经网络获取图像情绪数据；An image emotion analysis module for acquiring image emotion data according to the facial image information and the acquired deep convolutional neural network;

GPS模块，用于根据所述客户位置信息、环境信息和距离信息进 行三维重构，并获取避障路线。The GPS module is used to perform three-dimensional reconstruction according to the customer location information, environmental information and distance information, and obtain an obstacle avoidance route.

本申请实施例中提供的一个或多个技术方案，至少具有如下有益 效果：本申请实施例的机器人，包括设置有训练好的深度卷积神经网 络的动作分析模块、语音情绪分析模块和图像情绪分析模块，还包括 用于获取面部图像信息、动作信息、环境信息、客户位置信息、距离 信息的双目摄像头，能够为获取客户的满意度、服务人员的服务水平 和环境的干净整洁程度，通过多类型数据综合得出服务质量评价提供 了硬件基础；同时，还设置有GPS模块，能够获取避障路线，使得所 述机器人在不影响客户的前提下获取服务质量评价。One or more technical solutions provided in the embodiments of the present application have at least the following beneficial effects: the robot of the embodiments of the present application includes an action analysis module, a speech emotion analysis module and an image emotion provided with a trained deep convolutional neural network The analysis module also includes a binocular camera for obtaining facial image information, motion information, environmental information, customer location information, and distance information, which can be used to obtain customer satisfaction, service level of service personnel, and cleanliness of the environment. Multiple types of data are synthesized to obtain service quality evaluation, which provides a hardware basis; at the same time, a GPS module is also provided, which can obtain obstacle avoidance routes, so that the robot can obtain service quality evaluations without affecting customers.

第二方面，本申请还提供了一种服务质量评价方法，应用于上述 所述的一种机器人，包括以下步骤至少之一：In the second aspect, the application also provides a service quality evaluation method, applied to the above-mentioned robot, comprising at least one of the following steps:

所述机器人从服务器读取预先训练好的深度卷积神经网络；The robot reads the pre-trained deep convolutional neural network from the server;

所述机器人获取处于服务区域的客户和服务人员的语音情绪数 据和图像情绪数据并输入至所述深度卷积神经网络中，生成第一评分 数据并发送至所述服务器；The robot obtains voice emotion data and image emotion data of customers and service personnel in the service area and inputs them into the deep convolutional neural network, generates first scoring data and sends it to the server;

所述机器人获取环境信息并输入至所述深度卷积神经网络中，生 成第二评分数据并发送至所述服务器；The robot obtains environmental information and inputs it into the deep convolutional neural network, generates second scoring data and sends it to the server;

所述机器人检测到处于非服务区域下的客户的语音信息中包含 预先设定的关键词信息后，通过所述服务器向处于空闲状态的服务人 员发送提示信息并获取响应时长，根据所述响应时长生成第三评分数 据并发送至所述服务器；After the robot detects that the voice information of the customer in the non-service area contains preset keyword information, it sends prompt information to the service personnel in an idle state through the server and obtains the response time, according to the response time. generating third scoring data and sending it to the server;

所述机器人获取处于非服务区域的客户的动作数据并输入至所 述深度卷积神经网络中，生成第四评分数据并发送至所述服务器；The robot obtains the action data of the customer in the non-service area and inputs it into the deep convolutional neural network, generates the fourth scoring data and sends it to the server;

所述服务器根据所述第一评分数据、所述第二评分数据、所述第 三评分数据和所述第四评分数据生成服务质量评价数据。The server generates service quality evaluation data according to the first scoring data, the second scoring data, the third scoring data and the fourth scoring data.

本申请实施例中提供的一个或多个技术方案，至少具有如下有益 效果：本申请实施例的服务质量评价方法，应用于所述机器人，通过 所述机器人，获取服务区域内客户和服务人员的语音情绪数据和图像 情绪数据，能够得出客户的满意度和服务人员的服务水平作为第一评 分数据；根据所述环境信息能够得出环境整洁的程度作为第二评分数 据；根据服务人员在非服务区域内的响应速度生成第三评分数据；根 据非服务区域内的客户的动作数据能够得出客户对服务人员办事效 率的满意度生成第四评分数据，根据第一评分数据、第二评分数据、 第三评分数据和第四评分数据能够实现综合多类型数据得出服务质 量评价，使得服务质量评价更具有参考价值。One or more technical solutions provided in the embodiments of the present application have at least the following beneficial effects: the service quality evaluation method of the embodiments of the present application is applied to the robot, and the robot obtains the information of customers and service personnel in the service area through the robot. Voice emotion data and image emotion data can be used to obtain customer satisfaction and service level of service personnel as the first scoring data; according to the environmental information, the degree of cleanliness of the environment can be obtained as the second scoring data; The response speed in the service area generates the third score data; according to the action data of the customers in the non-service area, the customer's satisfaction with the service efficiency of the service personnel can be obtained to generate the fourth score data. According to the first score data and the second score data , the third scoring data and the fourth scoring data can realize the comprehensive multi-type data to obtain the service quality evaluation, which makes the service quality evaluation more valuable for reference.

进一步，若所述机器人检测到客户从服务区域移动至非服务区域， 还包括：获取所述客户的服务评价分数，并发送至服务器。Further, if the robot detects that the customer moves from the service area to the non-service area, the method further includes: acquiring the customer's service evaluation score and sending it to the server.

进一步，还包括：根据非服务区域内的面部图像信息和语音信息 得出服务人员的服务次数，将所述服务次数发送至所述服务器。Further, it also includes: obtaining the service times of the service personnel according to the facial image information and voice information in the non-service area, and sending the service times to the server.

进一步，所述通过所述服务器向处于空闲状态的服务人员发送提 示信息并获取响应时长后，还包括：Further, after the described server sends prompt information to the service personnel in the idle state and obtains the response duration, it also includes:

所述机器人获取客户位置信息、环境信息和距离信息并进行三维 重构，得出避障路线；The robot obtains customer location information, environmental information and distance information and performs three-dimensional reconstruction to obtain an obstacle avoidance route;

所述机器人根据所述避障路线移动，并根据所述关键词信息从服 务器读取出对应的操作信息并执行。The robot moves according to the obstacle avoidance route, and reads out the corresponding operation information from the server according to the keyword information and executes it.

进一步，还包括：Further, it also includes:

所述机器人获取非服务区域内处于空闲状态的服务人员的位置 信息；The robot obtains the location information of the service personnel in the idle state in the non-service area;

所述机器人根据所述处于空闲状态的服务人员的位置信息、所述 环境信息和所述距离信息进行三维重构，得出避障路线；The robot performs three-dimensional reconstruction according to the position information of the service personnel in the idle state, the environment information and the distance information, and obtains an obstacle avoidance route;

所述机器人根据所述避障路线移动，并读取测试交互信息，获取 所述处于空闲状态的服务人员的测试成绩，并发送至所述服务器。The robot moves according to the obstacle avoidance route, reads the test interaction information, obtains the test scores of the service personnel in the idle state, and sends them to the server.

进一步，还包括：Further, it also includes:

所述机器人根据所述语音情绪数据和动作数据获取所述深度卷 积神经网络的增量学习数据，并发送至服务器；The robot obtains the incremental learning data of the deep convolutional neural network according to the voice emotion data and the action data, and sends it to the server;

所述服务器根据所述增量学习数据，基于元学习的注意力吸引网 络结和循环式反向传播在所述深度卷积神经网络训练出新类别；According to the incremental learning data, the server attracts network nodes and cyclic back-propagation based on meta-learning attention to train a new category in the deep convolutional neural network;

所述服务器将更新后的所述深度卷积神经网络同步至所述机器 人中。The server synchronizes the updated deep convolutional neural network into the robot.

第三方面，本申请还提供了一种服务质量评价系统，包括机器人 群组和服务器，所述机器人群组由若干个如上述所述的一种机器人组 成，所述机器人群组和所述服务器配合执行如上述所述的服务质量评 价方法。In a third aspect, the present application also provides a service quality evaluation system, including a robot group and a server, the robot group is composed of several robots as described above, the robot group and the server Cooperate with the implementation of the service quality evaluation method described above.

第四方面，本申请提供了一种服务质量评价装置，包括至少一个 控制处理器和用于与所述至少一个控制处理器通信连接的存储器；所 述存储器存储有可被所述至少一个控制处理器执行的指令，所述指令 被所述至少一个控制处理器执行，以使所述至少一个控制处理器能够 执行如上述所述的服务质量评价方法。In a fourth aspect, the present application provides an apparatus for evaluating service quality, comprising at least one control processor and a memory for communicating with the at least one control processor; the memory stores data that can be processed by the at least one control processor. The instructions are executed by the at least one control processor to enable the at least one control processor to execute the quality of service evaluation method as described above.

第五方面，本申请提供了一种计算机可读存储介质，计算机可读 存储介质存储有计算机可执行指令，计算机可执行指令用于使计算机 执行如上所述的服务质量评价方法。In a fifth aspect, the present application provides a computer-readable storage medium, where the computer-readable storage medium stores computer-executable instructions, and the computer-executable instructions are used to cause a computer to execute the above-mentioned service quality evaluation method.

第六方面，本申请还提供了一种计算机程序产品，所述计算机程 序产品包括存储在计算机可读存储介质上的计算机程序，所述计算机 程序包括程序指令，当所述程序指令被计算机执行时，使计算机执行 如上所述的服务质量评价方法。In a sixth aspect, the present application also provides a computer program product, the computer program product includes a computer program stored on a computer-readable storage medium, the computer program includes program instructions, and when the program instructions are executed by a computer , and make the computer execute the above-mentioned service quality evaluation method.

附图说明Description of drawings

下面结合附图和实例对本申请作进一步说明。The present application will be further described below with reference to the accompanying drawings and examples.

图1是本申请一个实施例提供的一种机器人的模块示意图；1 is a schematic diagram of a module of a robot provided by an embodiment of the present application;

图2是本申请另一个实施例提供的一种服务质量评价方法的流 程图；Fig. 2 is a flow chart of a service quality evaluation method provided by another embodiment of the present application;

图3是本申请另一个实施例提供的一种服务质量评价方法的深 度卷积神经网络的结构示意图；Fig. 3 is the structural representation of the deep convolutional neural network of a kind of service quality evaluation method provided by another embodiment of the present application;

图4是本申请另一个实施例提供的一种服务质量评价方法的流 程图；Fig. 4 is a flow chart of a service quality evaluation method provided by another embodiment of the present application;

图5是本申请另一个实施例提供的一种服务质量评价方法的流 程图；Fig. 5 is a flow chart of a service quality evaluation method provided by another embodiment of the present application;

图6是本申请另一个实施例提供的一种服务质量评价方法的流 程图；Fig. 6 is a flow chart of a service quality evaluation method provided by another embodiment of the present application;

图7是本申请另一个实施例提供的一种服务质量评价方法的流 程图；Fig. 7 is a flow chart of a service quality evaluation method provided by another embodiment of the present application;

图8是本申请另一个实施例提供的一种服务质量评价方法的流 程图；Fig. 8 is a flow chart of a service quality evaluation method provided by another embodiment of the present application;

图9是本申请另一个实施例提供的一种服务质量评价系统的结 构示意图；9 is a schematic structural diagram of a service quality evaluation system provided by another embodiment of the present application;

图10是本申请第二实施例提供的一种用于执行服务质量评价方 法的装置示意图。Fig. 10 is a schematic diagram of an apparatus for executing a service quality evaluation method provided by the second embodiment of the present application.

具体实施方式Detailed ways

为了使本申请的目的、技术方案及优点更加清楚明白，以下结合 附图及实施例，对本申请进行进一步详细说明。应当理解，此处所描 述的具体实施例仅用以解释本申请，并不用于限定本申请。In order to make the purpose, technical solutions and advantages of the present application clearer, the present application will be described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application, but not to limit the present application.

需要说明的是，如果不冲突，本申请实施例中的各个特征可以相 互结合，均在本申请的保护范围之内。另外，虽然在装置示意图中进 行了功能模块划分，在流程图中示出了逻辑顺序，但是在某些情况下， 可以以不同于装置中的模块划分，或流程图中的顺序执行所示出或描 述的步骤。It should be noted that, if there is no conflict, various features in the embodiments of the present application can be combined with each other, which are all within the protection scope of the present application. In addition, although the functional modules are divided in the schematic diagram of the device, and the logical sequence is shown in the flowchart, in some cases, the modules in the device may be divided differently, or the sequence shown in the flowchart may be performed. or the described steps.

参考图1，本申请的实施例提供了一种机器人100，机器人100 包括：Referring to FIG. 1, an embodiment of the present application provides arobot 100. Therobot 100 includes:

双目摄像头110，用于获取面部图像信息、动作信息、环境信息、 客户位置信息、距离信息；Thebinocular camera 110 is used to obtain facial image information, motion information, environmental information, customer location information, and distance information;

语音接收模块120，用于接收语音信息；avoice receiving module 120, configured to receive voice information;

动作分析模块130，用于根据动作信息和所获取的深度卷积神经 网络获取动作数据；Action analysis module 130, for obtaining action data according to action information and the acquired deep convolutional neural network;

语音情绪分析模块140，用于根据语音信息和所获取的深度卷积 神经网络获取语音情绪数据；The speechemotion analysis module 140 is used for acquiring speech emotion data according to speech information and the acquired deep convolutional neural network;

图像情绪分析模块150，用于根据面部图像信息和所获取的深度 卷积神经网络获取图像情绪数据；The imageemotion analysis module 150 is used to obtain image emotion data according to the facial image information and the acquired deep convolutional neural network;

GPS模块160，用于根据客户位置信息、环境信息和距离信息进 行三维重构，并获取避障路线。TheGPS module 160 is used to perform three-dimensional reconstruction according to the customer's location information, environmental information and distance information, and obtain an obstacle avoidance route.

在一实施例中，双目摄像头110可以是现有技术中常见的任意型 号，在此不再赘述。需要说明的是，本实施例中的语音接收模块120 可以是现有技术中常见的拾音设备，例如是麦克风，本申请并不涉及 对语音接收模块120的硬件改进，在此不再赘述。需要说明的是，本 实施例中的动作分析模块130、语音情绪分析模块140和图像情绪分 析模块150为独立的功能模块，例如每个模块中配置单独的处理芯片， 该处理芯片能够接收输入的数据，并且运行深度卷积神经网络并得出 计算结果即可。需要说明的是，GPS模块160还可以包括常见的GPS 定位装置，能够获取机器人的当前位置，便于路径规划。本领域技术 人员可以理解的是，可以在GPS模块160中配置处理芯片，用于接收 双目摄像头110发送的数据并执行任意现有的三维重构算法，从而得 出避障路线即可。In one embodiment, thebinocular camera 110 may be any type commonly used in the prior art, and details are not described herein again. It should be noted that thevoice receiving module 120 in this embodiment can be a common sound pickup device in the prior art, such as a microphone, and the present application does not involve hardware improvements to thevoice receiving module 120, which will not be repeated here. It should be noted that theaction analysis module 130, the speechemotion analysis module 140 and the imageemotion analysis module 150 in this embodiment are independent functional modules, for example, each module is configured with a separate processing chip, and the processing chip can receive the input data, and run the deep convolutional neural network and get the calculation results. It should be noted that theGPS module 160 may also include a common GPS positioning device, which can acquire the current position of the robot, which is convenient for path planning. Those skilled in the art can understand that a processing chip may be configured in theGPS module 160 to receive data sent by thebinocular camera 110 and execute any existing three-dimensional reconstruction algorithm, thereby obtaining an obstacle avoidance route.

在一实施例中，该机器人100还可以包括近距离无线通信模块、 显示屏、控制按键等。其中，近距离无线通信模块又可以为WIFI模 块或者蓝牙模块；另外，当显示屏为触摸显示屏时，控制按键可以为 该触摸显示屏的一个按键功能。In one embodiment, therobot 100 may further include a short-range wireless communication module, a display screen, a control button, and the like. Wherein, the short-range wireless communication module may be a WIFI module or a Bluetooth module; in addition, when the display screen is a touch display screen, the control key may be a key function of the touch display screen.

在一实施例中，为了实现机器人100的移动，还可以设置驱动装 置，例如常见的驱动轮和转向轮，能够与GPS模块160电连接，并在 GPS模块的驱动作用下移动即可，在此不再赘述。In one embodiment, in order to realize the movement of therobot 100, a driving device, such as a common driving wheel and a steering wheel, can also be provided, which can be electrically connected to theGPS module 160 and move under the driving action of the GPS module. No longer.

参考图2，本申请的另一个实施例还提供了一种服务质量评价方 法，应用于上述的一种机器人，包括但不限于以下步骤至少之一：Referring to Fig. 2, another embodiment of the present application also provides a service quality evaluation method, which is applied to the above-mentioned robot, including but not limited to at least one of the following steps:

步骤S210，机器人从服务器读取预先训练好的深度卷积神经网 络；Step S210, the robot reads the pre-trained deep convolutional neural network from the server;

步骤S220，机器人获取处于服务区域的客户和服务人员的语音 情绪数据和图像情绪数据并输入至深度卷积神经网络中，生成第一评 分数据并发送至服务器；Step S220, the robot obtains the voice emotion data and image emotion data of the customers and service personnel in the service area and inputs them into the deep convolutional neural network, generates the first scoring data and sends it to the server;

步骤S230，机器人获取环境信息并输入至深度卷积神经网络中， 生成第二评分数据并发送至服务器；Step S230, the robot obtains the environmental information and inputs it into the deep convolutional neural network, generates second scoring data and sends it to the server;

步骤S240，机器人检测到处于非服务区域下的客户的语音信息 中包含预先设定的关键词信息后，通过服务器向处于空闲状态的服务 人员发送提示信息并获取响应时长，根据响应时长生成第三评分数据 并发送至服务器；Step S240, after the robot detects that the voice information of the customer in the non-service area contains preset keyword information, it sends prompt information to the service personnel in the idle state through the server and obtains the response time, and generates a third response time according to the response time. Scoring data and sending it to the server;

步骤S250，机器人获取处于非服务区域的客户的动作数据并输 入至深度卷积神经网络中，生成第四评分数据并发送至服务器；Step S250, the robot obtains the action data of the customer in the non-service area and inputs it into the deep convolutional neural network, generates the fourth scoring data and sends it to the server;

步骤S260，服务器根据第一评分数据、第二评分数据、第三评 分数据和第四评分数据生成服务质量评价数据。Step S260, the server generates service quality evaluation data according to the first scoring data, the second scoring data, the third scoring data and the fourth scoring data.

在一实施例中，由于深度卷积神经网络的训练需要较大量的计算， 本申请实施例优选通过服务器完成深度卷积神经网络的训练，机器人 在启动时从服务器读取并下载训练好的深度卷积神经网络即可。In one embodiment, since the training of the deep convolutional neural network requires a large amount of computation, the embodiment of the present application preferably completes the training of the deep convolutional neural network through the server, and the robot reads and downloads the trained depth from the server when it is started. Convolutional Neural Networks.

在一实施例中，为了便于评分，可以在服务器生成若干个评分账 号，例如为每个服务人员建立评分账号，并且评分数据中包括对应的 评分账号，可以通过根据对服务人员的面部识别结果查出，在此不再 赘述。In one embodiment, in order to facilitate the scoring, several scoring accounts can be generated on the server, for example, a scoring account is established for each service staff, and the scoring data includes the corresponding scoring account, which can be checked according to the facial recognition results of the service staff. out, and will not be repeated here.

在一实施例中，步骤S220中的第一评分数据可以是服务人员和 客户的态度，例如通过深度卷积神经网络对服务人员和客户的语音信 息，识别出该语音信息属于积极状态或者消极状态，再根据对应的状 态设置对应的分数即可，例如服务人员属于积极状态，则该服务人员 加一分，具体的评分标准根据实际需求调整即可，并不在本申请实施 例的改进范围内，在此不再赘述。In one embodiment, the first scoring data in step S220 may be the attitudes of service personnel and customers. For example, the voice information of the service personnel and customers is identified through a deep convolutional neural network to identify that the voice information belongs to a positive state or a negative state. , and then set the corresponding score according to the corresponding state. For example, if the service staff is in a positive state, the service staff will add one point. The specific scoring standard can be adjusted according to the actual needs, which is not within the scope of improvement of the embodiments of the present application. It is not repeated here.

在一实施例中，步骤S230中的第二评分数据可以是环境卫生的 分数，例如通过双目摄像头获取环境信息后输入至深度卷积神经网络， 通过与保存在服务器的默认环境图像进行对比，识别出环境信息中的 障碍物，例如默认环境图像中的地面为平整平面，而根据环境信息识 别出该平整平面上有若干个小体积物品，则可认为地面上有垃圾等， 在第二评分数据作出对应的扣分即可，也可以对一些固定摆放物品的 位置进行识别，以判断其是否被移位等，具体的评分标准并不在本实 施例的改进中，在此不再赘述。In one embodiment, the second scoring data in step S230 may be a score of environmental hygiene, such as obtaining environmental information through a binocular camera and then inputting it to a deep convolutional neural network, and comparing it with the default environmental image saved in the server, Obstacles in the environmental information are identified. For example, the ground in the default environmental image is a flat plane, and if there are several small-volume objects on the flat plane according to the environmental information, it can be considered that there is garbage on the ground, etc., in the second score The data can be deducted accordingly, and the positions of some fixed items can also be identified to determine whether they have been displaced.

在一实施例中，步骤S240中可以预先设置响应时长的阈值进行 判断，例如设置阈值为1分钟，若超过1分钟仍未检测到该客户处有 服务人员进行服务，则认定响应时长超期，在第三评分数据中进行相 应的分数调整即可。需要说明的是，当检测到关键词信息后，机器人 可以检测发出该关键词信息的客户附近是否有服务人员，若有，还可 以通过面部图像信息识别出具体的服务人员，并在第三评分数据中对 所对应的服务人员进行加分，具体的评分标准并不在本实施例的改进 中，在此不再赘述。In one embodiment, in step S240, a threshold value of the response duration can be preset for judgment, for example, the threshold value is set to 1 minute. If it is not detected that there is a service person at the customer for more than 1 minute, it is determined that the response duration is overdue. The corresponding score adjustment can be performed in the third scoring data. It should be noted that when the keyword information is detected, the robot can detect whether there is a service person near the customer who sent the keyword information. Points are added to the corresponding service personnel in the data, and the specific scoring standard is not included in the improvement of this embodiment, and will not be repeated here.

在一实施例中，步骤S250通过对客户动作数据的识别，能够用 于判断客户的情绪，例如通过深度卷积神经网络识别出客户在跺脚， 可以认定该客户处于不耐烦的情绪中，该不耐烦的情绪有可能是因为 服务效率较低引起的，在第四评分数据中对相应的服务人员进行一定 的分数调整即可。In one embodiment, step S250 can be used to judge the emotion of the customer by identifying the customer's action data. For example, it can be determined that the customer is in an impatient mood by recognizing that the customer is stomping through a deep convolutional neural network. The impatient mood may be caused by low service efficiency, and a certain score adjustment can be made to the corresponding service staff in the fourth scoring data.

在一实施例中，对客户的情绪判断之前，还可以判断该客户是否 处于与他人打电话等状态，例如识别出客户的动作为拿着手机等，能 够以通过面部图像识别和动作识别获取数据作出判断即可，在此不再 赘述。In one embodiment, before judging the customer's emotions, it is also possible to judge whether the customer is in a state of calling others, for example, it can be recognized that the customer's action is holding a mobile phone, etc., and data can be obtained through facial image recognition and motion recognition. It is enough to make a judgment, which is not repeated here.

基于上述实施例，第一评分数据、第二评分数据、第三评分数据、 第四评分数据中还优选包括与服务人员或服务部门对应的评分账号， 服务器接收到上述数据之后根据对应的评分账号完成分数的调整即 可，可以采用现有技术中任意的数据统计方法，在此不再赘述。Based on the foregoing embodiment, the first scoring data, the second scoring data, the third scoring data, and the fourth scoring data preferably also include scoring accounts corresponding to service personnel or service departments. After the server receives the above data, according to the corresponding scoring accounts It is only necessary to complete the adjustment of the score, and any data statistics method in the prior art may be adopted, which will not be repeated here.

参考图3，以下以一个具体实施例对本申请实施例中涉及的深度 卷积神经网络进行解释说明：Referring to Fig. 3 , the deep convolutional neural network involved in the embodiment of the present application is explained below with a specific embodiment:

在一实施例中，深度卷积神经网络是胶囊网络，包含一个由卷积 神经网络构成的编码器结构，一个卷积层用于提取初步的特征一个主 胶囊层用于接收卷积层检测到的基本特征生成特征组合，一个数字胶 囊层用于接收高级别的特征以及三层由全连接层组成的解码结构。编 码层将需要训练的数据包括：语音信息、面部图像信息等转换成低维 度的特征矩阵，并将矩阵传给主胶囊层进行特征的组合，由数字胶囊 层存储高级特征并在此期间通过动态路由的方式优化特征的提取。In one embodiment, the deep convolutional neural network is a capsule network, comprising an encoder structure consisting of a convolutional neural network, a convolutional layer for extracting preliminary features, and a main capsule layer for receiving convolutional layer detections. The basic features of the generated feature combination, a digital capsule layer is used to receive high-level features and three decoding structures consisting of fully connected layers. The coding layer converts the data to be trained, including voice information, facial image information, etc., into a low-dimensional feature matrix, and transmits the matrix to the main capsule layer for feature combination. The routing method optimizes feature extraction.

基于上述实施例，胶囊网络优选使用动态路由的方式更新权值， 高层胶囊的输出向量与底层胶囊的输出向量进行点积相乘，加上临时 权值参数，得到新的权值参数。通过点积相乘，可以得出高层胶囊与 底层胶囊的方向不同，进而使得最后更新的值也不同。例如，若高层 胶囊输出向量与底层输入向量两者相似，即两者向量方向小于九十度， 则权值被放大，同理，若两者不相似，则权值被缩小。迭代后会得到 一个路由权值系数的集合。并以此来优化特征提取的过程。Based on the above embodiment, the capsule network preferably uses dynamic routing to update the weights. The output vector of the high-level capsule is multiplied by the dot product of the output vector of the bottom-level capsule, and the temporary weight parameter is added to obtain a new weight parameter. By multiplying the dot products, it can be concluded that the directions of the high-level capsules and the bottom-level capsules are different, so that the final updated values are also different. For example, if the high-level capsule output vector is similar to the bottom-level input vector, that is, the two vector directions are less than ninety degrees, the weights are enlarged. Similarly, if the two are not similar, the weights are reduced. After iteration, a set of routing weight coefficients will be obtained. And use this to optimize the feature extraction process.

之后的解码模块从正确的数字胶囊中接受输出向量，并学习其编 码为需要的目标实例图像。解码器用来作为正则子，它接受正确的数 字胶囊的输出作为输入，重建一张标准像素大小的图像，损失函数为 重建图像和输入图像之间的欧式距离。解码器强制胶囊学习对重建原 始图像有用的特征，重建图像越接近输入图像则效果越好。所有在胶 囊之间传输的数据都是向量，向量保留了原特征中的方向与概率信息， 因此普通卷积层是无法进行这样的数据处理的。The subsequent decoding module accepts the output vector from the correct digital capsule and learns to encode it into the desired target instance image. The decoder is used as a regularizer, it accepts the output of the correct digital capsule as input, and reconstructs an image of standard pixel size. The loss function is the Euclidean distance between the reconstructed image and the input image. The decoder forces the capsule to learn features useful for reconstructing the original image, and the closer the reconstructed image is to the input image, the better. All data transmitted between capsules are vectors, which retain the direction and probability information in the original features, so ordinary convolutional layers cannot process such data.

本申请的另一个实施例还提供了一种服务质量评价方法，如图4 所示，图4是图2中步骤S220的细化流程的另一个实施例的示意图， 该步骤S220包括但不限于：Another embodiment of the present application further provides a service quality evaluation method, as shown in FIG. 4 , which is a schematic diagram of another embodiment of the refinement process of step S220 in FIG. 2 , where step S220 includes but is not limited to :

步骤S410，机器人获取客户的服务评价分数，并发送至服务器。In step S410, the robot obtains the customer's service evaluation score and sends it to the server.

在一实施例中，由于机器人也可以设置有显示屏，因此可以通过 显示屏与客户交互，从而得到客户的服务评价，可以采用现有技术中 常用的客户评价方法，在此不再赘述。In one embodiment, since the robot can also be provided with a display screen, it can interact with the customer through the display screen, thereby obtaining the customer's service evaluation, and the customer evaluation method commonly used in the prior art can be used, which will not be repeated here.

参考图5，本申请的另一个实施例还提供了一种服务质量评价方 法，包括但不限于以下步骤：Referring to Figure 5, another embodiment of the present application also provides a service quality evaluation method, including but not limited to the following steps:

步骤S510，机器人根据非服务区域内的面部图像信息和语音信 息得出服务人员的服务次数，将服务次数发送至服务器。Step S510, the robot obtains the service times of the service personnel according to the facial image information and voice information in the non-service area, and sends the service times to the server.

在一实施例中，步骤S510可以根据服务人员与客户之间的距离、 语音信息和面部表情信息判断二者是否发生交流，若是，则认定该服 务人员正在服务该客户，将服务次数加一，并且在服务器对该服务人 员进行一定的分数调整，在此不再赘述。In one embodiment, step S510 can judge whether the two communicate according to the distance between the service personnel and the customer, voice information and facial expression information. And a certain score adjustment is performed on the server on the server, which will not be repeated here.

本申请的另一个实施例还提供了一种服务质量评价方法，如图6 所示，图6是图2中步骤S240的细化流程的另一个实施例的示意图， 该步骤S240包括但不限于：Another embodiment of the present application further provides a service quality evaluation method, as shown in FIG. 6 , which is a schematic diagram of another embodiment of the refinement process of step S240 in FIG. 2 , where step S240 includes but is not limited to :

步骤S610，机器人获取客户位置信息、环境信息和距离信息并 进行三维重构，得出避障路线；Step S610, the robot obtains customer location information, environmental information and distance information and performs three-dimensional reconstruction to obtain an obstacle avoidance route;

步骤S620，机器人根据避障路线移动，并根据关键词信息从服 务器读取出对应的操作信息并执行。Step S620, the robot moves according to the obstacle avoidance route, and reads out the corresponding operation information from the server according to the keyword information and executes it.

在一实施例中，在通过服务器向服务人员发送提示信息后，本实 施例还优选通过控制机器人移动至客户所在区域，从而在服务人员到 达之前为客户提供服务，提高客户体验。具体的移动控制方法可以是 现有技术任意的控制方法，在此不再赘述。In one embodiment, after sending the prompt information to the service personnel through the server, this embodiment also preferably controls the robot to move to the area where the customer is located, so as to provide service for the customer before the arrival of the service personnel and improve the customer experience. The specific movement control method may be any control method in the prior art, which will not be repeated here.

在一实施例中，关键词信息可以在服务器预先设定，例如所能提 供的服务项目名称，可以根据该服务项目名称读取对应的客户服务操 作，例如通过显示屏提供对应的表格给客户填写等，具体的与客户交 互的方法并非本申请作出的改进，在此不再赘述。In one embodiment, the keyword information can be preset on the server, for example, the name of the service item that can be provided, and the corresponding customer service operation can be read according to the name of the service item, for example, a corresponding form can be provided through the display screen for the customer to fill out. etc., the specific method of interacting with the customer is not an improvement made by the present application, and will not be repeated here.

在一实施例中，三维重构可以采用现有技术中任意的三维重构方 式，在此不再赘述。In an embodiment, the three-dimensional reconstruction may adopt any three-dimensional reconstruction manner in the prior art, which will not be repeated here.

参考图7，在本申请的另一个实施例还提供了一种服务质量评价 方法，包括但不限于以下步骤：Referring to Figure 7, another embodiment of the present application also provides a service quality evaluation method, including but not limited to the following steps:

步骤S710，机器人获取非服务区域内处于空闲状态的服务人员 的位置信息；Step S710, the robot obtains the location information of the service personnel in the idle state in the non-service area;

步骤S720，机器人根据处于空闲状态的服务人员的位置信息、 环境信息和距离信息进行三维重构，得出避障路线；Step S720, the robot performs three-dimensional reconstruction according to the position information, environment information and distance information of the service personnel in the idle state, and obtains an obstacle avoidance route;

步骤S730，机器人根据避障路线移动，并读取测试交互信息， 获取处于空闲状态的服务人员的测试成绩，并发送至服务器。In step S730, the robot moves according to the obstacle avoidance route, reads the test interaction information, obtains the test scores of the service personnel in the idle state, and sends them to the server.

在一实施例中，空闲状态可以是服务人员一段时间内处于一个位 置并且未发生与客户交互的状态，在这种情况下，对服务人员进行测 试交互，能够有利于考核服务人员的专业水平，作为服务质量评价的 一个参考分数。可以理解的是，移动至服务人员所在区域后，可以通 过机器人中的显示屏显示预先设定的测试交互信息，例如是问答卷或 者服务场景模拟等，具体方法并非本申请所涉及的改进，能够实现类 似效果即可，在此不赘述。In one embodiment, the idle state may be a state in which the service personnel are in one position for a period of time and have not interacted with customers. In this case, testing and interacting with the service personnel can help assess the professional level of the service personnel. As a reference score for service quality evaluation. It can be understood that, after moving to the area where the service personnel are located, the preset test interaction information can be displayed on the display screen in the robot, such as question and answer questions or service scenario simulation, etc. The specific method is not an improvement involved in this application, and can A similar effect can be achieved, and details are not described here.

参考图8，本申请的另一个实施例还提供了一种服务质量评价方 法，还包括但不限于以下步骤：Referring to Fig. 8, another embodiment of the present application also provides a service quality evaluation method, which also includes but is not limited to the following steps:

步骤S810，机器人根据语音情绪数据和动作数据获取深度卷积 神经网络的增量学习数据，并发送至服务器；Step S810, the robot obtains the incremental learning data of the deep convolutional neural network according to the voice emotion data and the action data, and sends it to the server;

步骤S820，服务器根据增量学习数据，基于元学习的注意力吸 引网络结和循环式反向传播在深度卷积神经网络训练出新类别；Step S820, the server trains a new category in the deep convolutional neural network according to the incremental learning data, based on the attention of meta-learning to attract network knots and cyclic back-propagation;

步骤S830，服务器将更新后的深度卷积神经网络同步至机器人 中。Step S830, the server synchronizes the updated deep convolutional neural network to the robot.

在一实施例中，由于不同的客户的动作和语言习惯不同，本申请 实施例通过语音情绪数据和动作数据对深度卷积神经网络进行更新， 能够使得深度卷积神经网络的的识别更加精准，得出更好的识别效果。 需要说明的是，增量学习能够快速实现新类别的训练，也可以采用其 他训练方法，在此不再赘述。In one embodiment, since the actions and language habits of different customers are different, the embodiment of the present application updates the deep convolutional neural network through speech emotion data and action data, which can make the recognition of the deep convolutional neural network more accurate, get a better recognition effect. It should be noted that incremental learning can quickly realize the training of new categories, and other training methods can also be used, which will not be repeated here.

以下以一个具体示例对步骤S810至步骤S830进行举例说明：Steps S810 to S830 are described below with a specific example:

首先用传统的监督学习在固有类别上训练出一个分类器，且学习 到一个固定的特征表达此阶段为预训练阶段，在每个训练和测试的节 点，结合元学习正则化矩阵训练一个新类别分类器，将新增类别和固 有类别结合起来优化上一阶段的正则模块，使其能够在固有分类器之 后发挥作用。需要说明的是，在预训练阶段中没有额外的特殊操作， 给定固有类别分类所有的数据和对应的类别标签，训练一个固有的类 别分类器，及其对应的特征表达，得到一个基本的分类模型。First, a classifier is trained on the inherent category with traditional supervised learning, and a fixed feature expression is learned. This stage is the pre-training stage. At each training and testing node, a new category is trained by combining the meta-learning regularization matrix. The classifier, which combines the newly added category and the inherent category to optimize the regularization module of the previous stage, so that it can play a role after the inherent classifier. It should be noted that there is no additional special operation in the pre-training stage. Given an inherent category, classify all the data and the corresponding category label, train an inherent category classifier, and its corresponding feature expression to obtain a basic classification Model.

增量类别数据集D用于该阶段的少样本节点学习，对于每个学习 的N-K节点，每次都分别选中不同于固有类别的K个新增类，每个新 增类部分都有来自支撑集S的N张图片以及M张来自查询集Q的图片， 而S和Q两个图片集合，可以被当做每个节点学习时候用到的训练集 和验证集，每个节点都从训练集S学到一个新的分类器，同时分类器 对应的学习参数W只作用于本节点称为速用参数。为了能个衡量整体 的分类效果，训练算法过程中只允许接触新增类别的训练集S，验证 模型用新增类别加固有类别合并的验证集Q。The incremental category dataset D is used for the learning of few-sample nodes in this stage. For each N-K node learned, K new classes that are different from the inherent categories are selected each time, and each new class is partially supported by The N pictures of the set S and the M pictures from the query set Q, and the two picture sets S and Q can be used as the training set and verification set used by each node when learning, each node is from the training set S. A new classifier is learned, and the learning parameter W corresponding to the classifier only acts on this node, which is called a quick parameter. In order to measure the overall classification effect, only the training set S of the new category is allowed to be contacted during the training algorithm process, and the verification model uses the new category to reinforce the verification set Q with the category merge.

在一实施例中，可以在训练过程中使用元学习的正则约束，由于 元学习的整体过程就是迭代重复上一个阶段的训练过程，在新增训练 集上的到一个新的分类器，同时在验证集Q上进行性能验证，结合交 叉熵损失函数与额外引入的正则项R(θ)作为优化目标函数来学习更 新速用参数W，其中θ是元参数，结合嵌入到注意力吸引网络如下：

得到整 体的优化目标函数，模型参数W的本质就是优化新增类别的预测，则 针对局部的每个节点训练验证后的过程，可能导致固有类别的性能无 法保证，为了解决灾难性固有类别遗忘问题，本实施例使用注意力吸 引网络作为优化的正则项R，将固有类别的信息特征进行编码，之后 的参数化为恒用参数存储使用，并通过整个注意力吸引网络来最小化 学习参数θ如下：

当这个预 测类是

得到最小化 参数θ，其中正则项R(W,θ)是注意力吸引网络的核心点，公式如下：

其中

也就是注意力吸引网络的吸引力部分，W则是上述所述的权重参 数，通过基于马氏距离平方和外加一个偏置项，正则部分就可以实现 单一从新类别中获取学习信息的过程，避免灾难性遗忘问题。In one embodiment, the regular constraints of meta-learning can be used in the training process, because the overall process of meta-learning is to iteratively repeat the training process of the previous stage, and add a new classifier on the newly added training set, and at the same time The performance verification is performed on the validation set Q, and the cross-entropy loss function and the additionally introduced regular term R(θ) are used as the optimization objective function to learn and update the quick-use parameter W, where θ is a meta-parameter, which is embedded into the attention attracting network as follows:

The overall optimization objective function is obtained. The essence of the model parameter W is to optimize the prediction of the new category. The process of training and verification for each local node may lead to the inability to guarantee the performance of the inherent category. In order to solve the catastrophic inherent category forgetting problem , this embodiment uses the attention attracting network as the optimized regular term R, encodes the information features of the inherent category, and then parameterizes the constant parameter storage for use, and minimizes the learning parameter θ through the entire attention attracting network as follows :

When this predictor class is

The minimized parameter θ is obtained, where the regular term R(W, θ) is the core point of the attention attracting network. The formula is as follows:

in

That is, the attractive part of the attention attracting network, W is the weight parameter mentioned above, by adding a bias term based on the square sum of the Mahalanobis distance, the regular part can realize the process of obtaining learning information from a new category in a single way, avoiding Catastrophic forgetting problem.

参考图9，本申请的另一个实施例还提供了一种服务质量评价系 统900，包括机器人群组910和服务器920，机器人群组由若干个如 上述的一种机器人组成，机器人群组和服务器配合执行如上所述的服 务质量评价方法。Referring to FIG. 9, another embodiment of the present application further provides a servicequality evaluation system 900, including arobot group 910 and aserver 920, the robot group is composed of several robots as described above, the robot group and the server Cooperate with the implementation of the service quality evaluation method as described above.

以下以一个具体示例对本实施例的服务质量评价系统进行举例 说明：The service quality evaluation system of this embodiment is described below with a specific example:

在一实施例中，机器人群组中包括4个机器人，当4个机器人在 工作区域内启动后，服务器920将工作区域分为4个工作子区域，并 且根据4个机器人与工作子区域的距离就近分配成为第一机器人911、 第二机器人912、第三机器人913和第四机器人914，其中，第一机 器人911在办公窗口之间随机的监测窗口服务人员及客户，获取第一 评分数据，第二机器人912在工作区域内根据避障路线移动，并且对 地面的垃圾和桌子椅子上的污渍进行检测，获取第二评分数据，同时， 第二机器人912检测到客户从服务区域移动到非服务区域时，还可以 移动至该客户面前，并通过显示屏与客户互动，获取客户对服务过程 的评分；第三机器人913在非服务区域内待机，检测到关键词信息后， 例如客户的常用提问词，获取该客户的位置，并通过服务器920向处 于空闲状态的服务人员发送提示信息，并获取服务人员的响应时间， 生成第三评分数据，为了提高客户体验，发送提示信息后，第三机器 人913移动至该用户所处位置，并通过显示屏与客户进行交互，与此 同此，第三机器人913还可以在非服务区域内获取服务人员接待客户 或者引导客户的次数，例如通过语音信息识别判断，并将统计的次数 发送至服务器920中用于分数调整；第四机器人914在服务区域内的 排队客户进行监测，当监测到客户的动作数据异常，例如来回走动， 踱步等动作，生成第四评分数据并发送至服务器920中，能够服务人 员的工作效率进行考核，另外，与第三机器人913同理，第四机器人 914可以在服务区域内获取服务人员接待客户或者引导客户的次数， 在此不再赘述；此外上述机器人群组中的机器人还可以在检测到空闲 的服务人员后，移动至该服务人员面前并进行服务知识问答交互，将 问答成绩发送至服务器920中用于分数调整。In one embodiment, the robot group includes 4 robots. After the 4 robots are started in the working area, theserver 920 divides the working area into 4 working sub-areas, and according to the distance between the 4 robots and the working sub-area The nearest robot is assigned to become thefirst robot 911, thesecond robot 912, thethird robot 913 and thefourth robot 914, wherein thefirst robot 911 randomly monitors the service personnel and customers in the window between the office windows, obtains the first score data, and thefirst robot 911. Thesecond robot 912 moves according to the obstacle avoidance route in the working area, and detects the garbage on the ground and the stains on the tables and chairs to obtain the second scoring data. At the same time, thesecond robot 912 detects that the customer moves from the service area to the non-service area When the robot is activated, it can also move to the customer and interact with the customer through the display screen to obtain the customer's rating of the service process; thethird robot 913 waits in the non-service area and detects keyword information, such as the customer's common question words , obtain the location of the customer, and send prompt information to the service personnel in the idle state through theserver 920, and obtain the response time of the service personnel, and generate the third scoring data. In order to improve the customer experience, after sending the prompt information, thethird robot 913 Move to the location of the user, and interact with the customer through the display screen. At the same time, thethird robot 913 can also obtain the number of times the service staff has received customers or guided customers in the non-service area, for example, through voice information recognition and judgment , and send the counted times to theserver 920 for score adjustment; thefourth robot 914 monitors the queuing customers in the service area, and when it detects that the customer's motion data is abnormal, such as walking back and forth, pacing, etc. Four scoring data is sent to theserver 920, so that the work efficiency of the service personnel can be assessed. In addition, in the same way as thethird robot 913, thefourth robot 914 can obtain the number of times the service personnel received customers or guided customers in the service area. This is not repeated here; in addition, the robots in the above robot group can also move to the front of the service personnel after detecting the idle service personnel and perform service knowledge quiz interaction, and send the quiz results to theserver 920 for score adjustment.

参考图10，本申请的另一个实施例还提供了一种服务质量评价 装置1000，包括：存储器1010、控制处理器1020及存储在存储器 1020上并可在控制处理器1010上运行的计算机程序，控制处理器执 行所述计算机程序时实现如上任意实施例中的服务质量评价方法，例 如，执行以上描述的图2中的方法步骤S210至S260，图4中的方法 步骤S410，图5中的方法步骤S510,图6中的方法步骤S610至S620， 图7中的方法步骤S710至S730,图8中的方法步骤S810至S830。Referring to FIG. 10 , another embodiment of the present application further provides a servicequality evaluation apparatus 1000, including: amemory 1010, acontrol processor 1020, and a computer program stored in thememory 1020 and running on thecontrol processor 1010, When the control processor executes the computer program, the service quality evaluation method in any of the above embodiments is implemented, for example, the method steps S210 to S260 in FIG. 2 described above, the method step S410 in FIG. 4 , and the method in FIG. 5 are executed. Step S510 , the method steps S610 to S620 in FIG. 6 , the method steps S710 to S730 in FIG. 7 , and the method steps S810 to S830 in FIG. 8 .

控制处理器1020和存储器1010可以通过总线或者其他方式连接， 图10中以通过总线连接为例。Thecontrol processor 1020 and thememory 1010 may be connected by a bus or in other ways, and the connection by a bus is taken as an example in FIG. 10 .

存储器1010作为一种非暂态计算机可读存储介质，可用于存储 非暂态软件程序以及非暂态性计算机可执行程序。此外，存储器1010 可以包括高速随机存取存储器，还可以包括非暂态存储器，例如至少 一个磁盘存储器件、闪存器件或其他非暂态固态存储器件。在一些实 施方式中，存储器1010可选包括相对于控制处理器1020远程设置的 存储器，这些远程存储器可以通过网络连接至该服务质量评价装置 1000。上述网络的实例包括但不限于互联网、企业内部网、局域网、 移动通信网及其组合。Thememory 1010, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs and non-transitory computer-executable programs. Additionally,memory 1010 may include high-speed random access memory, and may also include non-transitory memory, such as at least one magnetic disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, thememory 1010 may optionally include memory located remotely from thecontrol processor 1020, and these remote memories may be connected to the quality ofservice assessment device 1000 via a network. Examples of such networks include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

以上所描述的装置实施例仅仅是示意性的，其中作为分离部件说 明的单元可以是或者也可以不是物理上分开的，即可以位于一个地方， 或者也可以分布到多个网络单元上。可以根据实际的需要选择其中的 部分或者全部模块来实现本实施例方案的目的。The apparatus embodiments described above are only illustrative, and the units described as separate components may or may not be physically separated, that is, may be located in one place, or may be distributed to multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution in this embodiment.

此外，本申请的另一个实施例还提供了一种计算机可读存储介质， 该计算机可读存储介质存储有计算机可执行指令，该计算机可执行指 令被一个或多个控制处理器执行，例如，被图10中的一个控制处理 器1020执行，可使得上述一个或多个控制处理器1020执行上述方法 实施例中的服务质量评价方法，例如，执行以上描述的图2中的方法 步骤S210至S260，图4中的方法步骤S410，图5中的方法步骤S510, 图6中的方法步骤S610至S620，图7中的方法步骤S710至S730, 图8中的方法步骤S810至S830。In addition, another embodiment of the present application also provides a computer-readable storage medium, where the computer-readable storage medium stores computer-executable instructions, and the computer-executable instructions are executed by one or more control processors, for example, Executed by onecontrol processor 1020 in FIG. 10, the above one ormore control processors 1020 can execute the service quality evaluation method in the above method embodiment, for example, perform the above-described method steps S210 to S260 in FIG. 2 , method step S410 in FIG. 4 , method step S510 in FIG. 5 , method steps S610 to S620 in FIG. 6 , method steps S710 to S730 in FIG. 7 , method steps S810 to S830 in FIG. 8 .

需要说明的是，由于本实施例中的用于执行服务质量评价方法的 装置与上述的服务质量评价方法基于相同的发明构思，因此，方法实 施例中的相应内容同样适用于本装置实施例，此处不再详述。It should be noted that, since the device for executing the service quality evaluation method in this embodiment is based on the same inventive concept as the above-mentioned service quality evaluation method, the corresponding content in the method embodiment is also applicable to this device embodiment. It will not be described in detail here.

通过以上的实施方式的描述，本领域技术人员可以清楚地了解到 各实施方式可借助软件加通用硬件平台的方式来实现。本领域技术人 员可以理解实现上述实施例方法中的全部或部分流程是可以通过计 算机程序来指令相关的硬件来完成，所述的程序可存储于计算机可读 取存储介质中，该程序在执行时，可包括如上述方法的实施例的流程。 其中，所述的存储介质可为磁碟、光盘、只读存储记忆体(ReadOnly Memory,ROM)或随机存储记忆体(Random Access Memory,RAM)等。From the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a general hardware platform. Those skilled in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing the relevant hardware through a computer program, and the program can be stored in a computer-readable storage medium, and when the program is executed , which may include the flow of an embodiment of the above-mentioned method. The storage medium may be a magnetic disk, an optical disk, a read-only memory (ReadOnly Memory, ROM), or a random access memory (Random Access Memory, RAM) or the like.

以上是对本申请的较佳实施进行了具体说明，但本申请并不局限 于上述实施方式，熟悉本领域的技术人员在不违背本申请精神的前提 下还可作出种种的等同变形或替换，这些等同的变形或替换均包含在 本申请权利要求所限定的范围内。The above is a specific description of the preferred implementation of the application, but the application is not limited to the above-mentioned embodiments. Those skilled in the art can also make various equivalent deformations or replacements on the premise of not violating the spirit of the application. These Equivalent modifications or substitutions are included within the scope defined by the claims of the present application.

Claims (10)

1. A robot, characterized in that the robot comprises:

the binocular camera is used for acquiring facial image information, action information, environment information, client position information and distance information;

the voice receiving module is used for receiving voice information;

the action analysis module is used for acquiring action data according to the action information and the acquired deep convolutional neural network;

the voice emotion analysis module is used for acquiring voice emotion data according to the voice information and the acquired deep convolutional neural network;

the image emotion analysis module is used for acquiring image emotion data according to the facial image information and the acquired deep convolutional neural network;

and the GPS module is used for carrying out three-dimensional reconstruction according to the client position information, the environment information and the distance information and acquiring an obstacle avoidance route.

2. A service quality evaluation method applied to a robot according to claim 1, comprising at least one of the following steps:

the robot reads a pre-trained deep convolutional neural network from a server;

the robot acquires voice emotion data and image emotion data of a client and a service worker in a service area, inputs the voice emotion data and the image emotion data into the deep convolutional neural network, generates first scoring data and sends the first scoring data to the server;

the robot acquires environmental information, inputs the environmental information into the deep convolutional neural network, generates second scoring data and sends the second scoring data to the server;

after detecting that voice information of a client in a non-service area contains preset keyword information, the robot sends prompt information to a service worker in an idle state through the server, acquires response time, generates third scoring data according to the response time and sends the third scoring data to the server;

the robot acquires action data of a client in a non-service area, inputs the action data into the deep convolutional neural network, generates fourth scoring data and sends the fourth scoring data to the server;

and the server generates service quality evaluation data according to the first grading data, the second grading data, the third grading data and the fourth grading data.

3. The method of claim 2, wherein if the robot detects that the customer moves from a service area to a non-service area, the method further comprises: and acquiring the service evaluation score of the client and sending the service evaluation score to a server.

4. The method of claim 2, further comprising: and obtaining the service times of service personnel according to the facial image information and the voice information in the non-service area, and sending the service times to the server.

5. The method for evaluating the service quality according to claim 2, wherein after the server sends the prompt message to the service personnel in the idle state and obtains the response time, the method further comprises the following steps:

the robot acquires client position information, environment information and distance information and carries out three-dimensional reconstruction to obtain an obstacle avoidance route;

and the robot moves according to the obstacle avoidance route, reads corresponding operation information from a server according to the keyword information and executes the operation information.

6. The method of claim 2, further comprising:

the robot acquires the position information of service personnel in an idle state in a non-service area;

the robot carries out three-dimensional reconstruction according to the position information of the service personnel in the idle state, the environment information and the distance information to obtain an obstacle avoidance route;

and the robot moves according to the obstacle avoidance route, reads the test interaction information, acquires the test result of the service personnel in the idle state, and sends the test result to the server.

7. The method of claim 2, further comprising:

the robot acquires incremental learning data of the deep convolutional neural network according to the voice emotion data and the action data and sends the incremental learning data to a server;

the server trains a new category in the deep convolutional neural network according to the incremental learning data based on the attention attracting network node and the cyclic back propagation of the meta learning;

the server synchronizes the updated deep convolutional neural network into the robot.

8. A service quality evaluation system characterized by: the method comprises a robot group and a server, wherein the robot group consists of a plurality of robots as claimed in claim 1, and the robot group and the server cooperate to execute the service quality evaluation method as claimed in any one of claims 2 to 7.

9. A quality of service evaluation apparatus comprising at least one control processor and a memory for communicative connection with the at least one control processor; the memory stores instructions executable by the at least one control processor to enable the at least one control processor to perform a quality of service assessment method according to any of claims 2 to 7.

10. A computer-readable storage medium characterized by: the computer-readable storage medium stores computer-executable instructions for causing a computer to perform the method of quality of service evaluation according to any one of claims 2 to 7.

Images