CN110377175B - Recognition method and system for tap event on touch panel, and terminal touch product - Google Patents
Recognition method and system for tap event on touch panel, and terminal touch productDownload PDFInfo
- Publication number
- CN110377175B CN110377175BCN201810330628.6ACN201810330628ACN110377175BCN 110377175 BCN110377175 BCN 110377175BCN 201810330628 ACN201810330628 ACN 201810330628ACN 110377175 BCN110377175 BCN 110377175B
- Authority
- CN
- China
- Prior art keywords
- touch
- tapping
- neural network
- network
- deep neural
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F3/00—Input arrangements for transferring data to be processed into a form capable of being handled by the computer; Output arrangements for transferring data from processing unit to output unit, e.g. interface arrangements
- G06F3/01—Input arrangements or combined input and output arrangements for interaction between user and computer
- G06F3/03—Arrangements for converting the position or the displacement of a member into a coded form
- G06F3/041—Digitisers, e.g. for touch screens or touch pads, characterised by the transducing means
- G06F3/0412—Digitisers structurally integrated in a display
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F3/00—Input arrangements for transferring data to be processed into a form capable of being handled by the computer; Output arrangements for transferring data from processing unit to output unit, e.g. interface arrangements
- G06F3/01—Input arrangements or combined input and output arrangements for interaction between user and computer
- G06F3/03—Arrangements for converting the position or the displacement of a member into a coded form
- G06F3/041—Digitisers, e.g. for touch screens or touch pads, characterised by the transducing means
- G06F3/0416—Control or interface arrangements specially adapted for digitisers
- G06F3/0418—Control or interface arrangements specially adapted for digitisers for error correction or compensation, e.g. based on parallax, calibration or alignment
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- General Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Human Computer Interaction (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Biophysics (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Computational Linguistics (AREA)
- Artificial Intelligence (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- User Interface Of Digital Computer (AREA)
- Electrophonic Musical Instruments (AREA)
Abstract
Description
Translated fromChinese技术领域technical field
本揭示涉及一种感测技术，特别有关一种触控面板上的敲击事件的识别方法及系统，以及终端触控产品。The disclosure relates to a sensing technology, in particular to a method and system for recognizing a tap event on a touch panel, and a terminal touch product.
背景技术Background technique
现有的大尺寸触控显示装置搭配有标注绘图软件，可供使用者在显示画面上进行标注，以方便解说画面上的内容。该标注绘图软件通常会在画面边缘显示主菜单，使用者可通过该主菜单改变画笔颜色或调整画笔粗细。然而，由于屏幕尺寸很大，该主菜单可能距离使用者很远，在点选上相当不便，调整画笔属性的操作对使用者来说相当麻烦。Existing large-size touch display devices are equipped with annotation drawing software for users to mark on the display screen to facilitate explanation of the content on the screen. The annotation drawing software usually displays a main menu on the edge of the screen, through which the user can change the color of the brush or adjust the thickness of the brush. However, due to the large size of the screen, the main menu may be far away from the user, and it is quite inconvenient to click, and the operation of adjusting the properties of the brush is quite troublesome for the user.
有鉴于此，有必要提出一种新的方案，以解决上述问题。In view of this, it is necessary to propose a new scheme to solve the above problems.
发明内容Contents of the invention
本发明的目的在于提供一种触控面板上的敲击事件的识别方法及系统，以及终端触控产品，以提升敲击类型之预测的准确度。The object of the present invention is to provide a method and system for recognizing a tap event on a touch panel, as well as a terminal touch product, so as to improve the accuracy of the prediction of the tap type.
为达成上述目的，本发明一方面提供一种触控面板上的敲击事件的识别方法，包含：感测在触控面板上进行的各种敲击事件而量测出的若干个振动信号，记录这些敲击事件的类型作为分类标记；将对应一个敲击事件的振动信号及分类标记作为一个样本，生成包含若干个样本的一样本集；将该样本集中的样本作为输入，自由选取的权重参数组作为调整参数，输入到一深度神经网络中进行训练，采用向后传播的算法，调整该权重参数组；将该样本集的样本分批读出，训练该深度神经网络，对该权重参数组进行微调，以决定出优化的权重参数组；将该深度神经网络及该优化的权重参数组作为一模型，布建到一终端触控产品中；感测对该终端触控产品执行的一敲击操作所产生的振动信号，并生成对应该敲击操作的若干个触碰感测值所形成的一影像；将对应该敲击操作的振动信号输入到该模型中以得出一第一结果，对这些触碰感测值所形成的影像进行分析以得出一第二结果，根据该第一结果和该第二结果得出一预测的敲击类型。In order to achieve the above object, the present invention provides a method for identifying a knock event on a touch panel, which includes: sensing a number of vibration signals measured from various knock events on the touch panel, Record the types of these knocking events as classification marks; use the vibration signal and classification marks corresponding to a knocking event as a sample to generate a sample set containing several samples; take the samples in the sample set as input, and freely select the weight The parameter group is used as an adjustment parameter, which is input into a deep neural network for training, and the weight parameter group is adjusted by using the algorithm of backward propagation; the samples of the sample set are read out in batches, and the deep neural network is trained, and the weight parameter Fine-tuning the group to determine the optimized weight parameter group; using the deep neural network and the optimized weight parameter group as a model, and deploying it in a terminal touch product; sensing a terminal touch product executed The vibration signal generated by the tapping operation, and an image formed by several touch sensing values corresponding to the tapping operation is generated; the vibration signal corresponding to the tapping operation is input into the model to obtain a first As a result, the image formed by the touch sensing values is analyzed to obtain a second result, and a predicted tap type is obtained according to the first result and the second result.
本发明另一方面提供一种触控面板上的敲击事件的识别方法，包含：感测在触控面板上进行的各种敲击事件而量测出的若干个振动信号，并将该触控面板感测每一敲击事件所得出的若干个触碰感测值转换成一影像，且记录这些敲击事件的类型作为分类标记；将对应一个敲击事件的振动信号、影像及分类标记作为一个样本，生成包含若干个样本的一样本集；将该样本集中的样本作为输入，自由选取的权重参数组作为调整参数，输入到一深度神经网络中进行训练，采用向后传播的算法，调整该权重参数组；以及将该样本集的样本分批读出，训练该深度神经网络，对该权重参数组进行微调，以决定出优化的权重参数组。Another aspect of the present invention provides a method for identifying a tap event on a touch panel, including: sensing several vibration signals measured from various tap events on the touch panel, and converting the touch The control panel detects each knocking event and converts several touch sensing values into an image, and records the type of these knocking events as a classification mark; the vibration signal, image and classification mark corresponding to a knocking event are used as One sample generates a sample set containing several samples; the samples in the sample set are used as input, and the freely selected weight parameter group is used as an adjustment parameter, which is input into a deep neural network for training, and the backward propagation algorithm is used to adjust the weight parameter set; and reading out the samples of the sample set in batches, training the deep neural network, and fine-tuning the weight parameter set to determine an optimized weight parameter set.
本发明再一方面提供一种触控面板上的敲击事件的识别系统，包含：一触控面板；一振动感测器，与该触控面板设置在一起，其中该振动感测器用以感测在触控面板上进行的各种敲击事件而量测出若干个振动信号，该触控面板用以感测每一敲击事件以得出的若干个触碰感测值，这些触碰感测值被转换成一影像；一处理器，与该振动感测器和该触控面板耦接，该处理器接收该振动感测器传来的这些振动信号以及该触控面板传来的若干张影像；以及一内存，与该处理器连接，该内存包含可由该处理器执行的若干个程序指令，该处理器执行这些程序指令以执行一方法，所述方法包含：将这些敲击事件的类型作为分类标记，记录在该内存中；将对应一个敲击事件的振动信号、影像及分类标记作为一个样本，生成包含若干个样本的一样本集；将该样本集中的样本作为输入，自由选取的权重参数组作为调整参数，输入到一深度神经网络中进行训练，采用向后传播的算法，调整该权重参数组；以及将该样本集的样本分批读出，训练该深度神经网络，对该权重参数组进行微调，以决定出优化的权重参数组。Yet another aspect of the present invention provides a system for recognizing a tap event on a touch panel, comprising: a touch panel; a vibration sensor set together with the touch panel, wherein the vibration sensor is used to sense Several vibration signals are measured by measuring various knock events performed on the touch panel. The touch panel is used to sense each knock event to obtain several touch sensing values. These touch The sensing value is converted into an image; a processor is coupled with the vibration sensor and the touch panel, and the processor receives the vibration signals from the vibration sensor and some vibration signals from the touch panel. an image; and a memory connected to the processor, the memory includes a number of program instructions executable by the processor, the processor executes these program instructions to perform a method, the method includes: combining the knock events The type is recorded in the memory as a classification mark; the vibration signal, image and classification mark corresponding to a knock event are used as a sample to generate a sample set containing several samples; the samples in the sample set are used as input and freely selected The weight parameter group is used as an adjustment parameter, and is input into a deep neural network for training, and the algorithm of backward propagation is used to adjust the weight parameter group; and the samples of the sample set are read out in batches, and the deep neural network is trained. The weight parameter set is fine-tuned to determine an optimized weight parameter set.
本发明又一方面提供一种终端触控产品，包含：一触控介面；一振动感测器，与该触控介面设置在一起，该振动感测器用以感测对该触控介面执行的一敲击操作而产生的振动信号，该触控介面用以感测该敲击操作以得出若干个触碰感测值，这些触碰感测值被转换成一影像；以及一控制器，与该振动感测器和该触控介面耦接，该控制器中布建有与上述方法中的深度神经网络对应的深度神经网络，该控制器用以将该对应的深度神经网络及根据上述方法得出的该优化的权重参数组作为一模型，并用以将对应该敲击操作的该振动信号及来自该触控介面的该影像输入该模型中，以得出一预测的敲击类型。Another aspect of the present invention provides a terminal touch product, including: a touch interface; a vibration sensor, which is set together with the touch interface, and the vibration sensor is used to sense the touch interface. A vibration signal generated by a tapping operation, the touch interface is used to sense the tapping operation to obtain a plurality of touch sensing values, and these touch sensing values are converted into an image; and a controller, and The vibration sensor is coupled to the touch interface, and a deep neural network corresponding to the deep neural network in the above method is deployed in the controller, and the controller is used to obtain the corresponding deep neural network and the above method. The optimized weight parameter set is used as a model, and is used to input the vibration signal corresponding to the tapping operation and the image from the touch interface into the model, so as to obtain a predicted tapping type.
本揭示采用深度学习的方式，运用深度神经网络学习在触控面板上进行的各种敲击事件的分类，得出一预测模型。将此预测模型布建在终端触控产品上，因此终端触控产品能够对用户作出的敲击动作进行预测，在软件层面上对这些敲击类型作不同的应用，使得可应用性大幅提高。并且，本揭示利用敲击产生的振动信号以及触碰感测值所形成的影像一起进行敲击类型的预测，使得敲击类型的预测准确度有效提升。This disclosure adopts the method of deep learning, uses the deep neural network to learn the classification of various tapping events on the touch panel, and obtains a prediction model. The predictive model is deployed on the terminal touch product, so the terminal touch product can predict the tap action made by the user, and apply these tap types differently at the software level, which greatly improves the applicability. Moreover, the present disclosure utilizes the vibration signal generated by the tapping and the image formed by the touch sensing value to predict the tapping type together, so that the prediction accuracy of the tapping type is effectively improved.
为让本揭示的上述内容能更明显易懂，下文特举优选实施例，并配合所附图式，作详细说明如下。In order to make the above content of the present disclosure more comprehensible, preferred embodiments are given below and described in detail in conjunction with the accompanying drawings.
附图说明Description of drawings
图1显示根据本揭示实现的一种触控面板上的敲击事件的识别系统的示意图。FIG. 1 shows a schematic diagram of a system for recognizing a tap event on a touch panel according to the present disclosure.
图2显示根据本揭示第一实施例实现的一种触控面板上的敲击事件的识别方法的流程图。FIG. 2 shows a flowchart of a method for recognizing a tap event on a touch panel according to the first embodiment of the present disclosure.
图3显示本揭示中的振动信号以时间分布的示意图。FIG. 3 shows a schematic diagram of the time distribution of vibration signals in the present disclosure.
图4显示本揭示中于频率空间的振动信号的示意图。FIG. 4 shows a schematic diagram of a vibration signal in frequency space in the present disclosure.
图5显示本揭示中的深度神经网络的示意图。FIG. 5 shows a schematic diagram of a deep neural network in the present disclosure.
图6显示根据本揭示第二实施例实现的一种触控面板上的敲击事件的识别方法的流程图。FIG. 6 shows a flowchart of a method for recognizing a tap event on a touch panel according to the second embodiment of the present disclosure.
图7显示根据本揭示的深度神经网络的架构示意图。FIG. 7 shows a schematic diagram of the architecture of a deep neural network according to the present disclosure.
图8显示根据本揭示实现的一种终端触控产品的示意图。FIG. 8 shows a schematic diagram of a terminal touch product implemented according to the present disclosure.
图9显示根据本揭示第二实施例实现的一种触控面板上的敲击事件的识别方法的流程图。FIG. 9 shows a flowchart of a method for recognizing a tap event on a touch panel according to the second embodiment of the present disclosure.
具体实施方式Detailed ways
为使本揭示的目的、技术方案及效果更加清楚、明确，以下参照附图并举实施例对本揭示进一步详细说明。应当理解，此处所描述的具体实施例仅用以解释本揭示，本揭示说明书所使用的词语“实施例”意指用作实例、示例或例证，并不用于限定本揭示。此外，本揭示说明书和所附权利要求书中所使用的冠词「一」一般地可以被解释为意指「一个或多个」，除非另外指定或从上下文可以清楚确定单数形式。In order to make the purpose, technical solutions and effects of the present disclosure more clear and definite, the present disclosure will be further described in detail below with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described here are only used to explain the present disclosure, and the word "embodiment" used in the present disclosure specification is meant to be used as an example, illustration or illustration, and is not intended to limit the present disclosure. Furthermore, the article "a" as used in this disclosure and the appended claims may generally be construed to mean "one or more" unless specified otherwise or clear from the context in the singular.
本揭示采用深度学习（deep learning）的方式，对在触控面板上作出的敲击事件进行分类学习而得出一分类模型。利用此分类模型，可以将使用者在触控产品上作出的敲击动作加以分类，得出敲击的类型（例如敲击的次数与敲击的手指数目的多种组合），从而可以执行对应该类型的预定操作。In this disclosure, a deep learning method is adopted to perform classification learning on the tap events made on the touch panel to obtain a classification model. Using this classification model, it is possible to classify the tap actions made by users on touch products, and obtain the types of taps (such as various combinations of the number of taps and the number of fingers tapped), so that the The scheduled action that should be typed.
特别地，本揭示利用振动感测器量测出的振动信号及触控面板量测出的触碰感测值的影像一起进行敲击类型的预测，能够有效提高敲击类型的预测准确度。In particular, the present disclosure utilizes the vibration signal measured by the vibration sensor and the image of the touch sensing value measured by the touch panel together to predict the tap type, which can effectively improve the prediction accuracy of the tap type.
敲击事件的类型例如可为手指不同部位进行的敲击（例如指腹敲击、指关节敲击、指甲敲击）、敲击的次数（例如一次敲击、二次敲击和三次敲击等）、同时敲击之手指的数量（例如一指敲击、二指敲击和三指敲击等）以及敲击的角度（例如90度敲击和45度敲击等），或上述多种类型的任意组合。The type of tap event can be, for example, taps performed on different parts of the finger (such as finger pad tap, knuckle tap, nail tap), the number of taps (such as one tap, two taps, and three taps) etc.), the number of fingers tapping at the same time (such as one-finger tapping, two-finger tapping and three-finger tapping, etc.), and the angle of tapping (such as 90-degree tapping and 45-degree tapping, etc.), or more Any combination of types.
该预定操作可根据不同的应用情境而有不同的配置。例如在大尺寸触控面板的应用情境下，举例来说，一指敲击可与开启/关闭主菜单的操作关联，二指敲击可与变更画笔颜色的操作关联，三指敲击可与变更画笔粗细的操作关联。如下文描述的，本领域技术人员可以理解，也可将本揭示的发明概念运用在其他应用。当然，敲击类型与执行之操作的关系也可由使用者自行定义。The predetermined operation may have different configurations according to different application scenarios. For example, in the application context of a large-size touch panel, for example, one-finger tap can be associated with the operation of opening/closing the main menu, two-finger tap can be associated with the operation of changing the brush color, and three-finger tap can be associated with Operation association for changing brush thickness. As will be described below, those skilled in the art will appreciate that the inventive concepts of the present disclosure may be employed in other applications as well. Of course, the relationship between the type of tapping and the operation to be performed can also be defined by the user.
图1显示根据本揭示实现的一种触控面板上的敲击事件的识别系统的示意图。该系统包括一触控装置10及与触控装置10耦接的计算机装置40。触控装置10包含一触控面板20及至少一振动感测器30。触控装置10可为具有触控功能的显示装置，其可通过一显示面板（未图示）显示影像，同时可接收使用者的触控操作。计算机装置40可为具有一定之运算能力的计算机，例如个人计算机、笔记型计算机等。本揭示中，为了对敲击事件进行分类，需要先收集敲击事件，在此，人为地敲击触控装置10，将敲击事件对应的信号传送到计算机装置40，计算机装置40采用深度神经网络（deep neural network）进行学习。FIG. 1 shows a schematic diagram of a system for recognizing a tap event on a touch panel according to the present disclosure. The system includes a
振动感测器30例如加速度计。振动感测器30可配置在触控装置10中的任一位置，较佳地，振动感测器30配置在触控面板20的下表面。振动感测器30用以感测在触控装置10上进行的敲击动作，产生相应的振动信号。当振动感测器30配置在触控面板20下表面时，可以针对在触控面板20上的敲击生成较好的信号。The
触控面板20包括一信号传送（Tx）层21及一信号接收（Rx）层22。触控面板20感测使用者的触摸操作，生成若干个触碰感测值（如电容值），从这些触碰感测值，可决定出触碰发生的位置。这些触碰感测值结合其座标（如X、Y座标）可视为一张影像，触碰感测值对应该影像中的像素值，此影像中触碰感测值的分布与敲击的形态是相应的。The
于一实施例中，计算机装置40通过接口接收振动感测器30生成的振动信号，将其馈入深度神经网络进行分类学习。人为产生敲击事件后，也可将每个敲击事件的类型作为分类标记（classification labels）输入到计算机装置40中，进行监督式学习（supervisedlearning）。In one embodiment, the computer device 40 receives the vibration signal generated by the
于另一实施例中，计算机装置40通过接口接收振动感测器30生成的振动信号以及来自触控面板20的触碰感测值，计算机装置40可将来自触控面板20的触碰感测值转换成影像的数据形式，而后将振动信号及该影像馈入深度神经网络进行分类学习。In another embodiment, the computer device 40 receives the vibration signal generated by the
也就是说，在进行敲击类型的预测之前，可将敲击事件产生的振动信号及对应该敲击事件之类型的分类标记作为输入来训练深度神经网络，也可将敲击事件产生的振动信号、影像信号（来自触控面板20）及对应该敲击事件之类型的分类标记作为输入来训练深度神经网络。That is to say, before the prediction of the type of tapping, the vibration signal generated by the tapping event and the classification mark corresponding to the type of the tapping event can be used as input to train the deep neural network, and the vibration signal generated by the tapping event can also be used The signal, the image signal (from the touch panel 20 ) and the classification label corresponding to the type of the tap event are used as input to train the deep neural network.
如图1所示，计算机装置40包含一处理器41及一内存42，处理器41与振动感测器30耦接，处理器41接收振动感测器30传来的振动信号，内存42与处理器41连接，内存42包含可由处理器41执行的若干个程序指令，处理器41执行这些程序指令以执行该深度神经网络的相关运算。计算机装置40也可利用GPU来执行该深度神经网络的相关运算，以提升运算速度。As shown in Figure 1, computer device 40 comprises a
图2显示根据本揭示第一实施例实现的一种触控面板上的敲击事件的识别方法的流程图。请配合图1，参阅图2，所述方法包括如下步骤。FIG. 2 shows a flowchart of a method for recognizing a tap event on a touch panel according to the first embodiment of the present disclosure. Please refer to FIG. 2 in conjunction with FIG. 1 , the method includes the following steps.
步骤S21：利用振动感测器30感测在触控面板20上进行的各种敲击事件而量测出的若干个振动信号，记录这些敲击事件的类型作为分类标记。在此步骤中，在触控面板20上人为地产生各种敲击事件，设置在触控面板20下表面上的振动感测器30感测敲击事件而生成振动信号。并且，对应每一敲击事件之类型的分类标记被记录下来，存储到内存42中。Step S21 : Use the
在本揭示中，振动感测器30的数量不限于一个，也可以是若干个，振动感测器30也可以设置在触控装置10中的任一位置，振动感测器30也可以感测在触控装置10表面的任一位置作出的敲击动作。In this disclosure, the number of
振动感测器30侦测到的加速度是时间函数，有三个方向分量，图3所示为某一敲击事件对应的加速度大小在时间上的分布。于一实施例中，可利用傅里叶转换分别将三个方向分量转换到频率空间，如图4所示。具体来说，所述方法可进一步包含将每一个振动信号从时间分布转换到频率空间的步骤。The acceleration detected by the
在转换到频率空间后，可进一步滤掉低频的直流成分（DC component）与高频杂讯，以避免分类结果受到重力加速度以及杂讯的影响。具体来说，所述方法可进一步包含对每一个振动信号进行滤波，过滤掉高频和低频的部分的步骤。After converting to the frequency space, the low-frequency DC component (DC component) and high-frequency noise can be further filtered out to avoid the classification results being affected by the acceleration of gravity and noise. Specifically, the method may further include the step of filtering each vibration signal to filter out high-frequency and low-frequency parts.
步骤S22：将对应一个敲击事件的振动信号及分类标记作为一个样本，生成包含若干个样本的一样本集。在此步骤中，振动感测器30量测的振动信号以及对应该敲击事件之类型的分类标记作为一笔数据，即一个样本，若干个样本构成一样本集。具体来说，一个样本包含一个振动信号的特征值（feature values）以及对应该敲击事件之类型的分类标记。Step S22: Taking the vibration signal and the classification mark corresponding to one tapping event as one sample, generating a sample set including several samples. In this step, the vibration signal measured by the
该样本集可分成训练样本集及测试样本集，该训练样本集可用来训练深度神经网络，该测试样本集用来测试训练得出的深度神经网络模型的分类准确度。The sample set can be divided into a training sample set and a test sample set. The training sample set can be used to train the deep neural network, and the test sample set is used to test the classification accuracy of the trained deep neural network model.
步骤S23：将该样本集中的样本作为输入，自由选取的权重参数组（weightingparameters）作为调整参数，输入到一深度神经网络中进行训练，采用向后传播（backpropagation）的算法，调整该权重参数组。Step S23: taking the samples in the sample set as input, and using freely selected weighting parameters as adjustment parameters, inputting them into a deep neural network for training, and using the backpropagation algorithm to adjust the weighting parameters .
步骤S22中的得出的一个样本中的特征值自输入层输入，通过该深度神经网络输出预测的分类标记。图5显示一个深度神经网络的例子，深度神经网络一般可以分为输入层、输出层及介于输入层和输出层间的学习层，样本集的每个样本从输入层输入，预测的分类标记从输出层输出。一般来说，深度神经网络包含许多学习层，其层数相当多（例如50～100层），故可实现深度学习。图5显示的深度神经网络仅为示意，本揭示的深度神经网络并不以此为限。The feature value in a sample obtained in step S22 is input from the input layer, and the predicted classification label is output through the deep neural network. Figure 5 shows an example of a deep neural network. A deep neural network can generally be divided into an input layer, an output layer, and a learning layer between the input layer and the output layer. Each sample of the sample set is input from the input layer, and the predicted classification mark output from the output layer. Generally speaking, a deep neural network contains many learning layers, and its number of layers is quite large (for example, 50-100 layers), so deep learning can be realized. The deep neural network shown in FIG. 5 is only for illustration, and the deep neural network disclosed in this disclosure is not limited thereto.
深度神经网络中可包含多个卷积层（convolutional layer）、批次归一层 （batchnormalization layer）、池化层 （pooling layer）、全连接层 （fully connected layer）、线性整流单元（rectified linear unit，ReLu）以及一个 Softmax 输出层等等。本揭示可以采用适当数量的层数进行学习，以在预测准确度与运算效率上取得平衡，但需注意的是，层数过多也可能导致准确度下降。深度神经网络可包含多个级联的子网络，每个子网络与位在其后的各个子网络相连，如DenseNet（Dense Convolutional Network），以提升预测的准确度。深度神经网络也可包含残留网络（Residual Network），用来解决降解（degradation）问题。A deep neural network can contain multiple convolutional layers, batch normalization layers, pooling layers, fully connected layers, rectified linear units , ReLu) and a Softmax output layer and so on. In this disclosure, an appropriate number of layers can be used for learning to achieve a balance between prediction accuracy and computational efficiency, but it should be noted that too many layers may also lead to a decrease in accuracy. A deep neural network can contain multiple cascaded sub-networks, and each sub-network is connected to subsequent sub-networks, such as DenseNet (Dense Convolutional Network), to improve prediction accuracy. The deep neural network can also contain a residual network (Residual Network) to solve the degradation problem.
该深度神经网络的目标是使得分类误差 (loss) 最小，优化的方法采用向后传播算法，也就是说，输出层得出的预测结果与真实的值进行比较，得到一个误差值，然后这个误差值逐层往回传，从而修正每一层的参数。The goal of the deep neural network is to minimize the classification error (loss). The optimization method uses the back propagation algorithm, that is, the prediction result obtained by the output layer is compared with the real value to obtain an error value, and then the error Values are passed back layer by layer, thereby modifying the parameters of each layer.
步骤S24：将该样本集的样本分批（mini-batch）读出，训练该深度神经网络，对该权重参数组进行微调，以决定出优化的权重参数组。每使用一批子样本集进行训练时，就会对权重参数组进行一次微调，如此迭代地进行，直到分类误差趋向于收敛。最后，选取出对于测试集有最高预测准确度的参数组作为优化的模型参数组。Step S24: Read out the samples of the sample set in batches (mini-batch), train the deep neural network, and fine-tune the weight parameter set to determine an optimized weight parameter set. Every time a batch of sub-sample sets are used for training, the weight parameter group will be fine-tuned once, and this is done iteratively until the classification error tends to converge. Finally, the parameter group with the highest prediction accuracy for the test set is selected as the optimized model parameter group.
步骤S25：将该深度神经网络及该优化的权重参数组作为一模型，布建到一终端触控产品中。在此步骤中，该终端触控产品具有一预测模型，其包含了与上述步骤S21～S24中采用的深度神经网络相同或相应的深度神经网络以及上述步骤S24得出的优化的权重参数组。Step S25: Deploy the deep neural network and the optimized weight parameter set as a model into a terminal touch product. In this step, the terminal touch product has a predictive model, which includes the same or corresponding deep neural network as the deep neural network used in the above steps S21-S24 and the optimized weight parameter set obtained in the above step S24.
步骤S26：感测对该终端触控产品执行的一敲击操作所产生的振动信号，并生成对应该敲击操作的若干个触碰感测值所形成的一影像。该终端触控产品包含有一触控介面及至少一振动感测器。在此步骤中，使用者敲击该终端触控产品时，该终端触控产品中的振动感测器量测出振动信号，该触控介面感测使用者的触摸操作，从而生成若干个触碰感测值（如电容值），这些触碰感测值被转换成一影像，此影像中触碰感测值的分布与敲击的形态是相应的。Step S26: Sensing a vibration signal generated by a tap operation performed on the terminal touch product, and generating an image formed by a plurality of touch sensing values corresponding to the tap operation. The terminal touch product includes a touch interface and at least one vibration sensor. In this step, when the user taps the terminal touch product, the vibration sensor in the terminal touch product measures the vibration signal, and the touch interface senses the user's touch operation, thereby generating several touch Touch sensing values (such as capacitance values), these touch sensing values are converted into an image, and the distribution of the touch sensing values in the image corresponds to the form of the tap.
步骤S27：将对应该敲击操作的振动信号输入到该模型中以得出一第一结果，对这些触碰感测值所形成的影像进行分析以得出一第二结果，根据该第一结果和该第二结果得出一预测的敲击类型。也就是说，敲击类型的预测是根据该终端触控产品之振动感测器产生的振动信号以及该终端触控产品之触控介面产生的触碰感测值所形成的影像而得出的。其中，来自该终端触控产品的振动信号作为该经过训练的深度神经网络的输入，以得出敲击类型的第一个预测结果。来自该终端触控产品的因触碰操作而形成的影像经过分析后得出敲击类型的第二个预测结果。该第一个预测结果和该第二个预测结果皆参与了最终之敲击类型的预测。Step S27: Input the vibration signal corresponding to the tapping operation into the model to obtain a first result, analyze the image formed by these touch sensing values to obtain a second result, according to the first The result and the second result yield a predicted tap type. That is to say, the prediction of the type of tap is based on the image formed by the vibration signal generated by the vibration sensor of the terminal touch product and the touch sensing value generated by the touch interface of the terminal touch product . Wherein, the vibration signal from the terminal touch product is used as the input of the trained deep neural network to obtain the first prediction result of the tapping type. The image formed by the touch operation from the terminal touch product is analyzed to obtain a second prediction result of the knock type. Both the first prediction result and the second prediction result participate in the final prediction of the tap type.
于一实施例中，假如该第一个预测结果和该第二个预测结果之其中一个预测结果为不确定或无法判定时，可由另一个预测结果决定最终之敲击类型的预测。于再一实施例中，最终预测的敲击类型可为该第一个预测结果和该第二个预测结果的其中一者，例如该第一个预测结果和该第二个预测结果不一致时，可预设选定其中一者作为最终的预测结果，或依其各自的得分来决定。于再一实施例中，最终预测的敲击类型可由该第一个预测结果和该第二个预测结果各别预测出的敲击类型的组合。In one embodiment, if one of the first prediction result and the second prediction result is uncertain or undecidable, the other prediction result may determine the final tap type prediction. In yet another embodiment, the final predicted tap type may be one of the first predicted result and the second predicted result, for example, when the first predicted result is inconsistent with the second predicted result, One of them can be selected by default as the final prediction result, or it can be determined according to their respective scores. In yet another embodiment, the final predicted tap type may be a combination of tap types respectively predicted by the first prediction result and the second prediction result.
举例来说，该模型从振动信号中可得出敲击的次数，例如二次敲击，从该影像中可以分析出是几根手指同时敲击，例如二指敲击，从而可以预测出敲击类型为进行了二次的二指敲击。举例来说，该模型从振动信号中可得出敲击的次数，例如二次敲击，从该影像中可以分析出手指哪一个部位进行了敲击，例如指关节敲击，从而可以预测出敲击类型为进行了二次的指关节敲击。For example, the model can obtain the number of taps from the vibration signal, such as the second tap, and from the image, it can be analyzed that several fingers are tapped at the same time, such as two-finger taps, so that the tap can be predicted. The tap type is a two-finger tap performed twice. For example, the model can get the number of taps from the vibration signal, such as the second tap. From this image, it can analyze which part of the finger is tapped, such as knuckle taps, so that it can predict The tap type is a knuckle tap performed twice.
步骤S28：根据该预测的敲击类型，执行对应该预测的敲击类型的一预定操作。在此步骤中，可将预测得出的敲击类型传送给执行中的一软件，该软件便执行对应预测结果的操作。Step S28: According to the predicted tap type, perform a predetermined operation corresponding to the predicted tap type. In this step, the predicted type of tap can be transmitted to a running software, and the software will perform an operation corresponding to the predicted result.
图6显示根据本揭示第二实施例实现的一种触控面板上的敲击事件的识别方法的流程图。请配合图1，参阅图6，所述方法包括如下步骤：FIG. 6 shows a flowchart of a method for recognizing a tap event on a touch panel according to the second embodiment of the present disclosure. Please cooperate with Fig. 1, referring to Fig. 6, described method comprises the following steps:
步骤S61：利用振动感测器30感测在触控面板20上进行的各种敲击事件而量测出的若干个振动信号，并将触控面板20感测每一敲击事件所得出的若干个触碰感测值转换成一影像，且记录这些敲击事件的类型作为分类标记。在此步骤中，在触控面板20上人为地产生各种敲击事件，设置在触控面板20下表面上的振动感测器30感测敲击事件而生成振动信号。触控面板20针对每一敲击事件产生若干个触碰感测值，这些触碰感测值被转换成一张影像的数据形式，一个敲击事件对应一张影像。并且，对应每一敲击事件之类型的分类标记被记录下来，存储到内存42中。Step S61: Utilize the
在本揭示中，触碰感测值的影像可由触控面板20产生后再输入到计算机装置40，也可由计算机装置40的处理器41来将来自触控面板20的触碰感测值转换成影像。In this disclosure, the image of the touch sensing value can be generated by the
通常，敲击事件只发生在触控面板20上的一个局部的区域，因此可以只撷取这个区域的触碰感测值的影像，也就是，可以只采用这个局部区域的影像进行深度神经网络的训练。具体来说，所述方法可进一步包含将包含该敲击事件之发生位置的一局部区域中的触碰感测值转换成该影像的步骤。Usually, the tap event only occurs in a local area on the
步骤S62：将对应一个敲击事件的振动信号、影像及分类标记作为一个样本，生成包含若干个样本的一样本集。在此步骤中，振动感测器30量测的振动信号、来自触控面板20的触碰感测值的影像以及对应该敲击事件之类型的分类标记作为一笔数据，即一个样本，若干个样本构成一样本集。具体来说，一个样本包含一个振动信号的特征值、该影像的特征值以及对应该敲击事件之类型的分类标记。Step S62: Taking the vibration signal, image and classification mark corresponding to a tapping event as a sample, generating a sample set including several samples. In this step, the vibration signal measured by the
该样本集可分成训练样本集及测试样本集，该训练样本集可用来训练深度神经网络，该测试样本集用来测试训练得出的深度神经网络模型的分类准确度。The sample set can be divided into a training sample set and a test sample set. The training sample set can be used to train the deep neural network, and the test sample set is used to test the classification accuracy of the trained deep neural network model.
步骤S63：将该样本集中的样本作为输入，自由选取的权重参数组作为调整参数，输入到一深度神经网络中进行训练，采用向后传播的算法，调整该权重参数组。Step S63: The samples in the sample set are used as input, and the freely selected weight parameter group is used as an adjustment parameter, which is input into a deep neural network for training, and the weight parameter group is adjusted by using a backpropagation algorithm.
图7显示根据本揭示的深度神经网络的架构示意图。在此步骤中，将该样本集中的每个样本输入到图7所示的深度神经网络进行训练。FIG. 7 shows a schematic diagram of the architecture of a deep neural network according to the present disclosure. In this step, each sample in the sample set is input to the deep neural network shown in Figure 7 for training.
如图7所示，该深度神经网络包含一第一子网络201、一第二子网络202、一平坦层（flatten layer）210及一多层感知器（Multi-Layer Perceptron）220。来自振动感测器30的振动信号输入第一子网络201，来自触控面板20的触碰感测值的影像输入第二子网络202。第一子网络201和第二子网络202在该深度神经网络中为平行的架构。振动信号为一维的数据，第一子网络201可采用一维的架构，例如一维的卷积神经网络（convolutionalneural network, CNN）。触碰感测值的影像为二维的数据，第二子网络202可采用二维的架构，例如二维的卷积神经网络。第一子网络201的输出和第二子网络202的输出，输入到平坦层210进行平坦化。由于第一子网络201为一维架构，第二子网络202为二维架构，需要利用平坦层210将第一子网络201和第二子网络202的输出转换成一维数据。平坦层210的输出再输入到多层感知器220进行敲击事件的分类预测，以得出一预测的分类标记。多层感知器220可视为该深度神经网络的另一个子网络，其由多个的节点层所组成，映射一组输入向量到一组输出向量，除了输入节点，每个节点都是一个带有非线性激活函数的神经元。多层感知器220可包含卷积层、ReLU层、池化层和全连接层。第一子网络201、第二子网络202和多层感知器220的神经元各对应一权重值，该深度神经网络的所有权重值构成该权重参数组。As shown in FIG. 7 , the deep neural network includes a
举例来说，多层感知器220由一个输入层，一或多个隐藏层及一个输出层组成，从平坦层210输入到多层感知器220的每一个输入值自输入层输入，每个输入值乘以一个权重值后求和（也可再求和后加上一偏差值（bias）），作为第一个隐藏层中一个神经元的值，这个值可再进行非线性转换得到最终的值，作为下一个隐藏层的输入，依次进行并在输出层输出预测结果，即预测的分类标记。For example, the
具体来说，步骤S63可进一步包含如下步骤：Specifically, step S63 may further include the following steps:
将对应一个敲击事件的振动信号的特征输入到一维的第一子网络201；Inputting the feature of the vibration signal corresponding to a knock event into the one-dimensional
将对应一个敲击事件之触碰感测值的影像的特征输入到二维的第二子网络202；Inputting the feature of the image of the touch sensing value corresponding to a knock event into the two-dimensional
将该第一子网络的输出和该第二子网络的输出，输入到平坦层210进行平坦化，以将该第一子网络201和该第二子网络202的输出转换成一维数据；以及The output of the first subnetwork and the output of the second subnetwork are input to the
将平坦层210的输出，输入到多层感知器220，进行敲击事件的分类预测，以得出预测的分类标记。The output of the
该深度神经网络的目标是使得分类误差 (loss) 最小，优化的方法采用向后传播算法，也就是说，输出层得出的预测结果与真实的值进行比较，得到一个误差值，然后这个误差值逐层往回传，从而修正每一层的参数。具体来说，所述方法可进一步包含根据该预测的分类标记与该样本中的分类标记的误差，采用该向后传播的算法，调整该权重参数组的步骤。该向后传播的算法不仅调整多层感知器220中各神经元的权重，也可调整第一子网络201和第二子网络202中各神经元的权重。The goal of the deep neural network is to minimize the classification error (loss). The optimization method uses the back propagation algorithm, that is, the prediction result obtained by the output layer is compared with the real value to obtain an error value, and then the error Values are passed back layer by layer, thereby modifying the parameters of each layer. Specifically, the method may further include a step of adjusting the weight parameter set according to the error between the predicted classification label and the classification label in the sample, using the backpropagation algorithm. The backward propagation algorithm not only adjusts the weights of the neurons in the
步骤S64：将该样本集的样本分批读出，训练该深度神经网络，对该权重参数组进行微调，以决定出优化的权重参数组。每使用一批子样本集进行训练时，就会对权重参数组进行一次微调，如此迭代地进行，直到分类误差趋向于收敛。最后，选取出对于测试集有最高预测准确度的参数组作为优化的模型参数组。Step S64: Read out the samples in the sample set in batches, train the deep neural network, and fine-tune the weight parameter set to determine an optimized weight parameter set. Every time a batch of sub-sample sets are used for training, the weight parameter group will be fine-tuned once, and this is done iteratively until the classification error tends to converge. Finally, the parameter group with the highest prediction accuracy for the test set is selected as the optimized model parameter group.
图8显示根据本揭示实现的一种终端触控产品的示意图。如图8所示，该终端触控产品包含一触控介面20’、一或多个振动感测器30’及一控制器60。振动感测器30’可以设置在触控介面20’的下表面，或者也可设置在该终端触控产品中的任一位置。振动感测器30’用以感测对该触控介面20’执行的一敲击操作而产生振动信号。触控介面20’用以感测该敲击操作以得出若干个触碰感测值，这些触碰感测值被转换成一影像。控制器60与触控介面20’和振动感测器30’耦接，接收触控介面20’产生的触碰感测值所形成之影像和振动感测器30’产生的振动信号。来自触控介面20’的触碰感测值亦可由控制器60转换成一张影像。FIG. 8 shows a schematic diagram of a terminal touch product implemented according to the present disclosure. As shown in FIG. 8 , the touch terminal product includes a touch interface 20', one or more vibration sensors 30' and a
控制器60用以对使用者在触控介面20’上进行的敲击事件进行分类预测，以得出一预测的敲击类型。举例来说，控制器60中布建有与上述步骤S61～S64中采用的深度神经网络相同或相应的深度神经网络，且存储有上述步骤S64得出的优化的权重参数组。该相应的深度神经网络及该优化的权重参数组构成一模型。控制器60将来自振动感测器30’的振动信号及来自触控介面20’之触碰感测值所形成的影像输入该模型中，即可得出预测的敲击类型。如此，该终端触控产品实现了敲击事件的分类预测。The
于一实施例中，控制器60可为该终端触控产品中的任一控制器。于另一实施例中，控制器60结合于一触控芯片中，也就是说，该终端触控产品的触控芯片不仅具有感测使用者的触摸操作的功能，同时也具有预测使用者的敲击类型的功能。具体来说，该深度神经网络相应的程序码以及该优化的权重参数组可写入触控芯片的固件中，在执行驱动程序的阶段，触控芯片可预测出敲击事件的类型。In one embodiment, the
请配合图6参阅图8，所述方法还包括如下步骤。Please refer to FIG. 8 in conjunction with FIG. 6 , the method further includes the following steps.
步骤S65：将该深度神经网络及该优化的权重参数组作为一模型，布建到一终端触控产品中。在此步骤中，该终端触控产品具有一预测模型，其包含了与上述步骤S61～S64中采用的深度神经网络相同或相应的深度神经网络以及上述步骤S64得出的优化的权重参数组。Step S65: Deploy the deep neural network and the optimized weight parameter set as a model into a terminal touch product. In this step, the terminal touch product has a predictive model, which includes the same or corresponding deep neural network as the deep neural network used in the above steps S61-S64 and the optimized weight parameter set obtained in the above step S64.
步骤S66：接收对该终端触控产品执行的一敲击操作所产生的振动信号及来自该终端触控产品之触碰感测值的影像，并将对应该敲击操作的该振动信号及来自该终端触控产品的该影像输入该模型中，以得出一预测的敲击类型。在此步骤中，使用者敲击该终端触控产品时，该终端触控产品中的振动感测器30’量测振动信号，触控介面20’产生触碰感测值，这些触碰感测值被转换成一影像，将该振动信号及该影像输入该模型中，即可预测得出敲击的类型。Step S66: Receive the vibration signal generated by a tap operation on the terminal touch product and the image of the touch sensing value from the terminal touch product, and send the vibration signal corresponding to the tap operation and the image from the The image of the terminal touch product is input into the model to obtain a predicted tap type. In this step, when the user taps the terminal touch product, the vibration sensor 30' in the terminal touch product measures the vibration signal, and the touch interface 20' generates touch sensing values. The measured value is converted into an image, and the vibration signal and the image are input into the model to predict the type of knock.
步骤S67：根据该预测的敲击类型，执行对应该预测的敲击类型的一预定操作。在此步骤中，控制器60可将预测得出的敲击类型传送给于操作系统中运作的一软件，该软件便执行对应预测结果的操作。Step S67: According to the predicted tap type, perform a predetermined operation corresponding to the predicted tap type. In this step, the
在一个例示的应用情境中，大尺寸触控显示产品中安装有一标注软件。举例来说，当使用者对此产品表面进行一指敲击时，该标记软件对应地开启/或关闭主菜单；二指敲击时，该标注软件改变画笔的颜色；三指敲击时，改变笔尖粗细。在另一个例示的应用情境中，当使用者进行90度敲击时，可开启/关闭主菜单，进行45度敲击时可以将菜单项目反白，供使用者选取多个项目，或者进行文字选取。在另一个例子中，当播放影片或音乐时，使用者从触控板侧面敲击一次可使播放暂停，敲击二次则可继续播放。In an exemplary application scenario, a labeling software is installed in a large-size touch display product. For example, when the user taps the surface of the product with one finger, the marking software correspondingly opens/or closes the main menu; when tapping with two fingers, the marking software changes the color of the brush; when tapping with three fingers, Change the thickness of the pen tip. In another exemplary application scenario, when the user taps at 90 degrees, the main menu can be opened/closed, and when the user taps at 45 degrees, the menu items can be highlighted for the user to select multiple items, or to enter text select. In another example, when a video or music is playing, the user taps once from the side of the touchpad to pause the playback, and taps twice to resume playback.
本揭示采用深度学习的方式，运用深度神经网络学习在触控面板上进行的各种敲击事件的分类，得出一预测模型。将此预测模型布建在终端触控产品上，因此终端触控产品能够对用户作出的敲击动作进行预测，在软件层面上对这些敲击类型作不同的应用，使得可应用性大幅提高。并且，本揭示利用敲击产生的振动信号以及触碰感测值所形成的影像一起进行敲击类型的预测，使得敲击类型的预测准确度有效提升。This disclosure adopts the method of deep learning, uses the deep neural network to learn the classification of various tapping events on the touch panel, and obtains a prediction model. The predictive model is deployed on the terminal touch product, so the terminal touch product can predict the tap action made by the user, and apply these tap types differently at the software level, which greatly improves the applicability. Moreover, the present disclosure utilizes the vibration signal generated by the tapping and the image formed by the touch sensing value to predict the tapping type together, so that the prediction accuracy of the tapping type is effectively improved.
本揭示已用较佳实施例揭露如上，然其并非用以限定本揭示，本揭示所属技术领域中具有通常知识者，在不脱离本揭示之精神和范围内，当可作各种之更动与润饰，因此本揭示之保护范围当视后附之权利要求书所界定者为准。This disclosure has been disclosed above with preferred embodiments, but it is not intended to limit this disclosure. Those with ordinary knowledge in the technical field to which this disclosure belongs can make various modifications without departing from the spirit and scope of this disclosure. and retouching, so the protection scope of this disclosure should be defined by the appended claims.
Claims (20)
Translated fromChinesePriority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810330628.6ACN110377175B (en) | 2018-04-13 | 2018-04-13 | Recognition method and system for tap event on touch panel, and terminal touch product |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810330628.6ACN110377175B (en) | 2018-04-13 | 2018-04-13 | Recognition method and system for tap event on touch panel, and terminal touch product |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110377175A CN110377175A (en) | 2019-10-25 |
| CN110377175Btrue CN110377175B (en) | 2023-02-03 |
Family
ID=68243332
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810330628.6AActiveCN110377175B (en) | 2018-04-13 | 2018-04-13 | Recognition method and system for tap event on touch panel, and terminal touch product |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110377175B (en) |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| TWI731442B (en)* | 2019-10-18 | 2021-06-21 | 宏碁股份有限公司 | Electronic apparatus and object information recognition method by using touch data thereof |
| US11620016B2 (en)* | 2020-07-30 | 2023-04-04 | Tdk Corporation | Information processing device and operation detection device |
| CN112148148B (en)* | 2020-09-27 | 2025-02-28 | 北京小米移动软件有限公司 | Touch operation recognition method, device, mobile terminal and storage medium |
| CN116027953A (en)* | 2022-08-15 | 2023-04-28 | 荣耀终端有限公司 | Recognition method for knuckle touch operation, electronic device and readable storage medium |
| CN116450026B (en)* | 2023-06-16 | 2023-10-20 | 荣耀终端有限公司 | Method and system for identifying touch operation |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CA2804014A1 (en)* | 2010-06-28 | 2012-01-12 | Cleankeys Inc. | Method for detecting and locating keypress-events on touch-and vibration-sensitive flat surfaces |
| CN105431799A (en)* | 2013-08-02 | 2016-03-23 | 齐科斯欧公司 | Capture vibroacoustic data used to determine touch type |
| CN106779064A (en)* | 2016-11-25 | 2017-05-31 | 电子科技大学 | Deep neural network self-training method based on data characteristics |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106021364B (en)* | 2016-05-10 | 2017-12-12 | 百度在线网络技术（北京）有限公司 | Foundation, image searching method and the device of picture searching dependency prediction model |
- 2018
- 2018-04-13CNCN201810330628.6Apatent/CN110377175B/enactiveActive
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CA2804014A1 (en)* | 2010-06-28 | 2012-01-12 | Cleankeys Inc. | Method for detecting and locating keypress-events on touch-and vibration-sensitive flat surfaces |
| CN105431799A (en)* | 2013-08-02 | 2016-03-23 | 齐科斯欧公司 | Capture vibroacoustic data used to determine touch type |
| CN106779064A (en)* | 2016-11-25 | 2017-05-31 | 电子科技大学 | Deep neural network self-training method based on data characteristics |
Non-Patent Citations (1)
| Title |
|---|
| 面向社会安全事件的分布式神经网络攻击行为分类方法;肖圣龙等;《计算机应用》;20171010(第10期);全文* |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110377175A (en) | 2019-10-25 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| TWI654541B (en) | Method and system for identifying tapping events on a touch panel, and terminal touch products | |
| CN110377175B (en) | Recognition method and system for tap event on touch panel, and terminal touch product | |
| Ling et al. | Ultragesture: Fine-grained gesture sensing and recognition | |
| TW201918866A (en) | Method and system for classifying tap events on touch panel, and touch panel product | |
| Zhang et al. | Human activity recognition across scenes and categories based on CSI | |
| CN102687161B (en) | Selective motor control classification | |
| Zhang et al. | Toward Deep Neural Networks: WASD Neuronet Models, Algorithms, and Applications | |
| Jiang et al. | Detection of underwater acoustic target using beamforming and neural network in shallow water | |
| KR102533084B1 (en) | Apparatus and method for processing imbalance data | |
| CN114091611A (en) | Device load weight acquisition method, device, storage medium and electronic device | |
| CN107729144B (en) | Application control method and device, storage medium and electronic equipment | |
| Sarakon et al. | Multisource Data Fusion Using MLP for Human Activity Recognition. | |
| Chen et al. | Tactile-GAT: tactile graph attention networks for robot tactile perception classification | |
| TWM595256U (en) | Intelligent gesture recognition device | |
| Ren et al. | Real-time continuous gesture recognition system based on PSO-PNN | |
| Yadav et al. | Gesture Recognition System for Human-Computer Interaction using Computer Vision | |
| CN109753172A (en) | Classification method and system of touch panel tap event, and touch panel product | |
| US12340627B2 (en) | System and methods for gesture inference using computer vision | |
| US12366923B2 (en) | Systems and methods for gesture inference using ML model selection | |
| US20170178032A1 (en) | Using a generic classifier to train a personalized classifier for wearable devices | |
| Amin et al. | Towards a trade-off between accuracy and computational cost for embedded systems: A tactile sensing system for object classification | |
| JP6947460B1 (en) | Programs, information processing equipment, and methods | |
| US12366920B2 (en) | Systems and methods for gesture inference using transformations | |
| CN111797852A (en) | Algorithm model adjustment method, device, storage medium and electronic device | |
| CN111813639B (en) | Method and device for evaluating equipment operation level, storage medium and electronic equipment |
Legal Events
| 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 |