CN112668609A - Tactile modal identification method based on kernel method - Google Patents
Tactile modal identification method based on kernel method Download PDFInfo
- Publication number
- CN112668609A CN112668609A CN202011416980.5A CN202011416980A CN112668609A CN 112668609 A CN112668609 A CN 112668609A CN 202011416980 A CN202011416980 A CN 202011416980A CN 112668609 A CN112668609 A CN 112668609A
- Authority
- CN
- China
- Prior art keywords
- kernel
- kernel function
- tensor
- regularized
- ideal
- 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.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 59
- 239000002131 composite material Substances 0.000 claims abstract description 37
- 238000000354 decomposition reaction Methods 0.000 claims abstract description 21
- 230000009467 reduction Effects 0.000 claims abstract description 10
- 238000012706 support-vector machine Methods 0.000 claims abstract description 8
- 238000012549 training Methods 0.000 claims description 15
- 239000011159 matrix material Substances 0.000 claims description 12
- 238000000513 principal component analysis Methods 0.000 claims description 5
- 238000005070 sampling Methods 0.000 claims description 4
- 102000008297 Nuclear Matrix-Associated Proteins Human genes 0.000 claims description 3
- 108010035916 Nuclear Matrix-Associated Proteins Proteins 0.000 claims description 3
- 210000000299 nuclear matrix Anatomy 0.000 claims description 3
- 239000004576 sand Substances 0.000 claims description 3
- 239000000126 substance Substances 0.000 claims description 3
- 238000013459 approach Methods 0.000 claims description 2
- 230000006870 function Effects 0.000 abstract description 56
- 230000003993 interaction Effects 0.000 description 8
- 238000012545 processing Methods 0.000 description 5
- 238000002474 experimental method Methods 0.000 description 4
- 241000282414 Homo sapiens Species 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 3
- 238000011161 development Methods 0.000 description 3
- 238000005259 measurement Methods 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 238000004422 calculation algorithm Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 230000008909 emotion recognition Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 239000000463 material Substances 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000008447 perception Effects 0.000 description 2
- 230000001133 acceleration Effects 0.000 description 1
- 230000009471 action Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000003592 biomimetic effect Effects 0.000 description 1
- 230000008451 emotion Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 230000035807 sensation Effects 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
Images
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a haptic modal identification method based on a kernel method, which belongs to the technical field of intelligent touch of robots, adopts a classification method of a composite kernel function, has higher learning efficiency and higher classification accuracy, and comprises the following steps: 1) acquiring a data set of tensor tactile signals; 2) performing dimensionality reduction treatment; 3) establishing a kernel function based on singular value decomposition; 4) establishing a global alignment kernel function; 5) establishing an ideal regularized composite kernel function based on a kernel function of singular value decomposition and a global alignment kernel function; 6) and fusing the ideal regularized composite kernel function with a support vector machine to obtain a kernel classification learner based on the tensor tactile signal, namely performing tactile modal identification on the tensor tactile signal.
Description
Technical Field
The invention belongs to the technical field of intelligent touch of robots, and particularly relates to a touch mode identification method based on a kernel method.
Background
In the human-computer interaction process, it is an important ability that the robot can timely sense and interact information transmitted by human beings, and especially, it is important that the robot can correctly sense and recognize tactile signals from the outside. The research on intelligent touch of the robot is gradually emphasized by people, and three technical problems are mainly involved in touch perception: high dimensionality of the haptic signal, complex tensor morphology of the haptic perception unit, and the problem of skew between different haptic time series samples, respectively. Indeed, in addition to this, various challenges are faced in the diverse task of haptic recognition, such as the prevalence of misalignment between haptic measurements, unequal durations of haptic measurements, inconsistent start time points for haptic measurements, and so forth.
As is well known, touch is a very important ability for people to perceive changes in the outside world, but in conventional industrial robots, touch sensors that are not generally sensitive to the robot are not given unless they perform some particular activity. In recent years, the field of intelligent touch of robots has attracted more and more attention, and touch interaction among human-computer interaction is slowly becoming popular, and under the current scene of human-computer interaction, touch has become the basic capability of interaction between robots and human. For example, many biomimetic robots are also widely used in different industries.
The related technology discloses an emotion recognition device and method based on an array type touch sensor, wherein the array type touch sensor is placed on a carrier and connected with a microcontroller, and is used for collecting pressure value data of a participant executing action and sending the pressure value data to the microcontroller; the microcontroller is connected with the upper computer and used for controlling the array type touch sensors to work, receiving and storing pressure value data collected by the array type touch sensors and then sending the pressure value data to the upper computer; and the upper computer stores the pressure value data received from the microcontroller, and displays the emotion recognition result in real time after analysis and processing.
The related technology also discloses a texture image cross-modal retrieval method based on the tactile texture features, which sequentially comprises the steps of selecting tactile texture training sample materials, establishing a tactile texture training data set and a texture image training data set, extracting the features of tactile acceleration and the like. The method utilizes the friction vibration signal of the collected texture surface as the tactile feature of the texture surface to retrieve the texture surface image most similar to the retrieved surface from the texture image retrieval library, namely, the cross-mode object material retrieval based on the tactile feature is realized.
However, in the related art, from the viewpoint of classification, there is only a few research on recognizing a touch manner by using a touch sensor, and some classification is considered by sensing a touch by using the touch sensor, determining an emotion carried by the touch, and the like; from the used classifiers, learning classifiers are usually simple and traditional classifiers, for example, the extracted main features are in a Support Vector Machine (SVM), a nearest neighbor node algorithm (KNN), and the like, and the classification learning device of the related art is too simple, low in learning efficiency, low in classification accuracy, and difficult to significantly improve.
Disclosure of Invention
In order to solve the problems in the prior art, the invention provides a haptic mode identification method based on a kernel method.
In order to achieve the above object, the present invention provides a haptic modality recognition method based on a kernel method, including:
1) acquiring a data set of tensor tactile signals;
2) performing dimensionality reduction treatment;
3) establishing a kernel function based on singular value decomposition;
4) establishing a global alignment kernel function;
5) establishing an ideal regularized composite kernel function based on a kernel function of singular value decomposition and a global alignment kernel function;
6) and fusing the ideal regularized composite kernel function with a support vector machine to obtain a kernel classification learner based on the tensor tactile signal, namely performing tactile modal identification on the tensor tactile signal.
Further, the dimensionality reduction processing is carried out in the step 2) by adopting downsampling and sparse principal component analysis.
Further, the step 3) comprises:
3.1) for haptic tensor signalsXIs subjected to matrix expansion to obtain X(n)To X(n)Singular value decomposition is performed as:
3.2) for tensorX iAndX jthe obtained kernel function is:
wherein | · | purple sweetFIs the Frobenius norm, σ is a hyperparameter;
3.3) the kernel function based on singular value decomposition is:
wherein N is the number of dimensions.
Further, the step 4) is for tensorX iAndX jthe global alignment kernel of (a) is as follows:
wherein the content of the first and second substances,
is all possible alignment paths, ψ (exp (Φ)), ψ (x, y) — | | | x-y | | luminance2S is
p is the length of the alignment path, t is the sampling time, phi is the kernel function, pi is a certain alignment path, pi1Is composed ofX iAssociated alignment path, pi2Is composed ofX jThe associated alignment path.
Further, the step 5) ideally regularized composite kernel function is expressed as:
KCK(X i,X j)＝(1-μ)KSVD(X i,X j)+μKGAK(X i,X j)
where μ is a trade-off factor that balances the singular value decomposition based kernel function and the global alignment kernel function.
Further, if the ideal regularized composite kernel incorporates label information, the ideal regularized composite kernel is expressed as:
wherein, K is a nuclear matrix,
in order to minimize the trace of the matrix, gamma is a super parameter, tr is the trace of the matrix, and T is an ideal kernel matrix;
through solving, the optimal K value is obtained as follows:
wherein "" indicates the product at the level of two model elements.
Further, the ideal regularized composite kernel is further generalized to present new samplesX sAndX tthen, the ideally regularized composite kernel function is expressed as:
wherein N istrainIs the number of training samples; and S ═ KCK)-1(KCK*+KCK)(KCK)-1。
Further, the step 6) includes: given training set
Wherein, yiIs a class label, and the prediction function of the kernel classification learner is expressed as:
wherein alpha isiAre weight coefficients.
Compared with the prior art, the method based on the ideal regularized composite kernel function has higher classification accuracy, has remarkable performance compared with many traditional learning methods, is too simple, has low learning efficiency and low classification accuracy, is difficult to obviously improve, provides a thought for fusing and creating a new classification learner by the classification method of the composite kernel function, uses the idea of fusing kernels, and simultaneously experiments show that the composite kernel function has higher learning efficiency, higher classification accuracy and more clear research prospect, and is expected to have greater development in the fields of artificial intelligence and machine learning.
Drawings
FIG. 1 is a flow chart of the method of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all, embodiments of the present invention. Thus, the following detailed description of the embodiments of the present invention is not intended to limit the scope of the invention as claimed, but is merely representative of selected embodiments of the invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The invention provides a tactile modality recognition method based on a kernel method, which can be applied to a human-computer interaction scene of robot tactile sensation with high sensitivity, and the robot can quickly recognize tactile signals of a person and respond, so that the robot has larger imagination space and development potential in the personification direction of an intelligent robot.
Referring to fig. 1, the present invention includes:
1) acquiring a data set of tensor tactile signals;
2) performing dimensionality reduction treatment;
3) establishing a kernel function based on singular value decomposition;
4) establishing a global alignment kernel function;
5) establishing an ideal regularized composite kernel function based on a kernel function of singular value decomposition and a global alignment kernel function;
6) and fusing the ideal regularized composite kernel function with a support vector machine to obtain a kernel classification learner based on the tensor tactile signal, namely performing tactile modal identification on the tensor tactile signal.
The method specifically comprises the following steps:
1) firstly, a data set of tensor tactile signals is collected through an open network channel;
2) carrying out dimensionality reduction on the data set of the tensor tactile signal by adopting downsampling and Sparse Principal Component Analysis (SPCA); the haptic signal is high-dimensional in nature, and because of the dimensionality disaster, it is impractical to feed the original haptic signal into the classical machine learning method;
3) establishing a Singular Value Decomposition (SVD) based kernel function:
3.1) for haptic tensor signalsXIs subjected to matrix expansion to obtain X(n)To X(n)Singular value decomposition is performed as:
3.2) for tensorX iAndX jthe obtained kernel function is:
wherein | · | purple sweetFIs the Frobenius norm, σ is a hyperparameter;
3.3) the kernel function based on singular value decomposition is:
wherein N is a dimension;
4) establishing a global alignment kernel function:
for tensorX iAndX jthe global alignment kernel of (a) is as follows:
wherein the content of the first and second substances,
is all possible alignment paths, ψ (exp (Φ)), ψ (x, y) — | | | x-y | | luminance2S is
p is the length of the alignment path, t is the sampling time, phi is the kernel function, pi is a certain alignment path, pi1Is composed ofX iAssociated alignment path, pi2Is composed ofX jAn associated alignment path; the calculation of the global alignment kernel function has quadratic calculation complexity which is almost the same as that of a dynamic time adjustment algorithm (DTW);
5) building an ideal regularized composite kernel function:
5.1) a composite kernel is a kernel obtained by fusing a plurality of kernels, in order to use the discrimination function of each kernel, specifically, a composite kernel should be given a weight corresponding to the sub-kernels constituting it, and the ideally regularized composite kernel in the present invention is a kernel obtained by fusing a Singular Value Decomposition (SVD) -based kernel and a global alignment kernel, and the ideally regularized composite kernel is expressed as follows:
wherein μ is a trade-off factor that balances the singular value decomposition based kernel function and the global alignment kernel function;
5.2) generally, the conventional composite kernel method does not take the label information into account, so in order to obtain a satisfactory composite kernel function by learning, the label information is fused into an ideal regularized composite kernel function, and the ideal regularized composite kernel function is expressed as follows:
wherein, K is a nuclear matrix,
in order to minimize the trace of the matrix, gamma is a super parameter, tr is the trace of the matrix, and T is an ideal kernel matrix;
through solving, the optimal K value is obtained as follows:
wherein, an indicates the product on the two model element levels;
5.3) the ideally regularized complex kernel is further generalized to give rise to new samplesX sAndX tthen, the ideally regularized composite kernel function is expressed as:
wherein N istrainIs the number of training samples; and S ═ KCK)-1(KCK*+KCK)(KCK)-1；
6) And (3) fusing an ideal regularized composite kernel function with a support vector machine to obtain a kernel classification learner based on tensor tactile signals, namely performing tactile modal identification on the tensor tactile signals: first, a training set is given
Wherein, yiIs a classification label, and then according to the expression theorem, the prediction function of the kernel classification learner is expressed as:
wherein alpha isiAre weight coefficients.
Firstly, a data set of tensor tactile signals is collected through a public network channel, and dimensionality reduction processing is carried out on tensor data with high dimensionality by combining a Sparse Principal Component Analysis (SPCA) and a downsampling method; the invention adopts the global alignment kernel function to solve the time dislocation problem of the haptic time sequence, ensures the unity of the haptic time, can improve the classification accuracy, integrates the advantages of kernels based on Singular Value Decomposition (SVD) and the global alignment kernel function to create an ideal regularized composite kernel function method, and simultaneously considers the label information of a training set, performs dimension reduction processing on tensor haptic signals, simplifies the complex tensor signals of a haptic sensing unit, solves the problem of deviation of the haptic time, and improves the classification accuracy of a system for identifying a haptic mode from the outside, and the invention can adopt three classifiers: based on a kernel function method of Singular Value Decomposition (SVD), a global alignment kernel function method and an ideal regularization composite kernel function method, the three methods are fused with a support vector machine to obtain three kinds of classification learners, and meanwhile, the three kinds of classification learners have good performance in accuracy.
The invention provides an open thought, provides a method for fusing various kernel functions to create an ideal regularized composite kernel function, can classify touch modes more correctly, and improves the classification accuracy. Therefore, the method based on the kernel and various classification learning methods are combined, and some new creations can be created in the field of man-machine intelligent interaction.
In the invention, two methods of Sparse Principal Component Analysis (SPCA) and downsampling are comprehensively used for feature extraction in the feature extraction stage, and the complex high-dimension tensor tactile signals are subjected to dimension reduction processing, so that the dimension disaster is avoided, the data is extracted more efficiently, the calculation steps are simplified, the complexity of a subsequent training model is reduced, and the classification efficiency is also improved.
The ideal regularized composite kernel function method finally obtained by the invention is compared with most classification learning methods, and the classification learning device has the best performance and the highest classification accuracy on the final classification accuracy, and is obviously higher than other classification learning devices. The reason is that it not only fuses the discriminant capabilities of the sub-kernel functions, but also takes into account the label information of the training set.
The invention has been proved to be correct through experiments and simulations. The training set data used in the experiment is a public data set, so the training set data can be queried through a network approach. The method has strong haptic mode recognition capability through a large amount of data training, and the test results of most of the classification learners are compared in experiments, so that the classification learning effect of the method is obviously superior to that of other classification learners, the classification accuracy of the method is higher than that of other methods, and the classification result is better improved compared with the previous classification result, and therefore, the method is proved to be feasible. The method is applied to a human-computer interaction scene requiring high-sensitivity robot touch, the robot can quickly identify human touch signals and respond, and the method has a larger imagination space and development potential in the personification direction of the intelligent robot.
The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention, and various modifications and changes may be made by those skilled in the art. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (8)
1. A haptic modality recognition method based on a kernel method, characterized by comprising:
1) acquiring a data set of tensor tactile signals;
2) performing dimensionality reduction treatment;
3) establishing a kernel function based on singular value decomposition;
4) establishing a global alignment kernel function;
5) establishing an ideal regularized composite kernel function based on a kernel function of singular value decomposition and a global alignment kernel function;
6) and fusing the ideal regularized composite kernel function with a support vector machine to obtain a kernel classification learner based on the tensor tactile signal, namely performing tactile modal identification on the tensor tactile signal.
2. The method for recognizing tactile modality based on the nuclear method according to claim 1, wherein the step 2) adopts down-sampling and sparse principal component analysis for dimension reduction.
3. A method of haptic modality recognition based on a nuclear method according to claim 1, characterized in that the step 3) comprises:
3.1) for haptic tensor signalsXIs subjected to matrix expansion to obtain X(n)To X(n)Singular value decomposition is performed as:
3.2) for tensorX iAndX jthe obtained kernel function is:
wherein | · | purple sweetFIs the Frobenius norm, σ is a hyperparameter;
3.3) the kernel function based on singular value decomposition is:
wherein N is the number of dimensions.
4. A method of haptic modality identification based on a nuclear method according to claim 1, characterized in that said step 4) is for tensorX iAndX jthe global alignment kernel of (a) is as follows:
wherein the content of the first and second substances,
is all possible alignment paths, ψ (exp (Φ)), ψ (x, y) — | | | x-y | | luminance2S is
p is the length of the alignment path, t is the sampling time, phi is the kernel function, not a certain alignment path, pi1Is composed ofX iAssociated alignment path, pi2Is composed ofX jThe associated alignment path.
5. A haptic modality recognition method based on a kernel method according to claim 1, characterized in that the ideal regularized composite kernel function of the step 5) is expressed as:
KCK(X i，X j)＝(1-μ)KSVD(X i，X j)+μKGAK(X i，X j)
where μ is a trade-off factor that balances the singular value decomposition based kernel function and the global alignment kernel function.
6. A haptic modality recognition method based on a kernel method according to claim 5, characterized in that the ideal regularized composite kernel function incorporates label information, and then the ideal regularized composite kernel function is expressed as:
wherein, K is a nuclear matrix,
in order to minimize the trace of the matrix, gamma is a super parameter, tr is the trace of the matrix, and T is an ideal kernel matrix;
through solving, the optimal K value is obtained as follows:
KCK*＝exp(log KCK+γT)
＝KCK⊙exp(γT)
wherein "" indicates the product at the level of two model elements.
7. A method for haptic modality recognition based on kernel-based approach as claimed in claim 6, wherein the ideal regularized composite kernel is further generalized to present new samplesX sAndX tthen, the ideally regularized composite kernel function is expressed as:
wherein N istrainIs the number of training samples; and S ═ KCK)-1(KCK*+KCK)(KCK)-1。
8. A method of haptic modality recognition based on a nuclear method according to claim 1, characterized in that the step 6) comprises: given training set
Wherein, yiIs a class label, and the prediction function of the kernel classification learner is expressed as:
wherein alpha isiAre weight coefficients.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011416980.5A CN112668609A (en) | 2020-12-07 | 2020-12-07 | Tactile modal identification method based on kernel method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011416980.5A CN112668609A (en) | 2020-12-07 | 2020-12-07 | Tactile modal identification method based on kernel method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112668609A true CN112668609A (en) | 2021-04-16 |
Family
ID=75401328
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011416980.5A Pending CN112668609A (en) | 2020-12-07 | 2020-12-07 | Tactile modal identification method based on kernel method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112668609A (en) |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20050265607A1 (en) * | 2004-05-13 | 2005-12-01 | Proximex | Multimodal high-dimensional data fusion for classification and identification |
| CN105005787A (en) * | 2015-06-24 | 2015-10-28 | 清华大学 | Dexterous hand tactile information based material classification method based on joint sparse coding |
| CN105956351A (en) * | 2016-07-05 | 2016-09-21 | 上海航天控制技术研究所 | Touch information classified computing and modelling method based on machine learning |
| CN106250855A (en) * | 2016-08-02 | 2016-12-21 | 南京邮电大学 | A kind of multi-modal emotion identification method based on Multiple Kernel Learning |
| CN110346142A (en) * | 2019-07-25 | 2019-10-18 | 哈尔滨理工大学 | Fault Diagnosis of Roller Bearings under varying load based on the alignment of unsupervised feature |
| CN111920420A (en) * | 2020-07-28 | 2020-11-13 | 复旦大学 | Patient behavior multi-modal analysis and prediction system based on statistical learning |
-
2020
- 2020-12-07 CN CN202011416980.5A patent/CN112668609A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20050265607A1 (en) * | 2004-05-13 | 2005-12-01 | Proximex | Multimodal high-dimensional data fusion for classification and identification |
| CN105005787A (en) * | 2015-06-24 | 2015-10-28 | 清华大学 | Dexterous hand tactile information based material classification method based on joint sparse coding |
| CN105956351A (en) * | 2016-07-05 | 2016-09-21 | 上海航天控制技术研究所 | Touch information classified computing and modelling method based on machine learning |
| CN106250855A (en) * | 2016-08-02 | 2016-12-21 | 南京邮电大学 | A kind of multi-modal emotion identification method based on Multiple Kernel Learning |
| CN110346142A (en) * | 2019-07-25 | 2019-10-18 | 哈尔滨理工大学 | Fault Diagnosis of Roller Bearings under varying load based on the alignment of unsupervised feature |
| CN111920420A (en) * | 2020-07-28 | 2020-11-13 | 复旦大学 | Patient behavior multi-modal analysis and prediction system based on statistical learning |
Non-Patent Citations (4)
| Title |
|---|
| PAOLO GASTALDO 等: ""A Tensor-Based Pattern-Recognition Framework for the Interpretation of Touch Modality in Artificial Skin Systems"", 《IEEE SENSORS JOURNAL》, vol. 14, no. 7, 29 April 2014 (2014-04-29) * |
| 方璐;徐丽蕊;董言治;: "基于机器学习的多用户触觉手势识别研究", 电声技术, no. 07, 5 July 2018 (2018-07-05) * |
| 李彦琳: ""基于深度学习和足底触觉特征的步态分析研究"", 《中国优秀硕士学位论文全文数据库 (信息科技辑)》, 15 July 2019 (2019-07-15) * |
| 梁栋;高玮玮;张艳;鲍文霞;: "基于足底压力图像的静态触觉步态识别", 华中科技大学学报(自然科学版), no. 10, 23 October 2013 (2013-10-23) * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110069129B (en) | Determination system and determination method | |
| CN112541529A (en) | Expression and posture fusion bimodal teaching evaluation method, device and storage medium | |
| Benalcázar et al. | Real-time hand gesture recognition based on artificial feed-forward neural networks and EMG | |
| Kumar et al. | Static hand gesture recognition using stacked denoising sparse autoencoders | |
| Yang et al. | Image classification using kernel collaborative representation with regularized least square | |
| Alameh et al. | DCNN for tactile sensory data classification based on transfer learning | |
| Ahmad et al. | Multidomain multimodal fusion for human action recognition using inertial sensors | |
| Hussain et al. | Intelligent sign language recognition system for e-learning context | |
| Minhas et al. | A fast recognition framework based on extreme learning machine using hybrid object information | |
| Alon et al. | Deep-hand: a deep inference vision approach of recognizing a hand sign language using american alphabet | |
| Pattar et al. | Intention and engagement recognition for personalized human-robot interaction, an integrated and deep learning approach | |
| Richardson et al. | Learning to predict perceptual distributions of haptic adjectives | |
| Ullah et al. | Emotion recognition from occluded facial images using deep ensemble model | |
| Barbhuiya et al. | Alexnet-CNN based feature extraction and classification of multiclass ASL hand gestures | |
| Boncolmo et al. | Gender Identification Using Keras Model Through Detection of Face | |
| Shah et al. | Real-time facial emotion recognition | |
| CN112668609A (en) | Tactile modal identification method based on kernel method | |
| CN115689981A (en) | Lung image detection method and device based on information fusion and storage medium | |
| Mukhanov et al. | GESTURE RECOGNITION OF MACHINE LEARNING AND CONVOLUTIONAL NEURAL NETWORK METHODS FOR KAZAKH SIGN LANGUAGE | |
| Boddu et al. | Novel heuristic recurrent neural network framework to handle automatic telugu text categorization from handwritten text image | |
| Zim | OpenCV and Python for Emotion Analysis of Face Expressions | |
| Riaz et al. | Surface EMG Real-Time Chinese Language Recognition Using Artificial Neural Networks | |
| Abdualghani et al. | Face expression recognition in grayscale images using image segmentation and deep learning | |
| Chivate et al. | Comparative Analysis of Multiple ML Models and Real-Time Translation | |
| Bhavatarini et al. | Enhancing Hand Gesture Recognition with Media Pipe and SVM Model on Custom Dataset |
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 |