Acoustic signal-based arm lifting gesture two-factor identity authentication method
Through ultrasonic recognition of arm lift gestures and combining deep learning models for identity authentication, the security risks and equipment cost problems of card swiping behavior authentication in the existing technology are solved, and more hidden, stable and accurate identity authentication is achieved.
Patent Information
- Application Number
- CN202411104489.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-13
- Publication Date
- 2024-11-22
- Estimated Expiration
- 2044-08-13
AI Technical Summary
The verification of card swiping behavior of existing smart cards poses a security threat, especially the recognition method of arm lifting movements based on wearable devices and cameras is easily affected by environmental factors and is not easy to promote, and lacks concealment and stability.
The two-factor identity authentication method of arm lift gesture based on sound signals is adopted. The ultrasonic transceiver device recognizes the arm lift gesture when swiping the card through an ultrasonic transceiver device. Combined with the improved LSTM-FCN model and SVM classifier, the arm lift action signal is extracted and judged to achieve identity authentication.
It improves the security and accuracy of card swiping behavior, reduces the impact of equipment costs and environmental factors, provides a hidden and easy-to-promote two-factor identification method, and enhances user information security.
Smart Images

Figure CN119004057A_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to the field of information security and Internet of Things security, and relates to a double-factor identity authentication method based on an arm-raising gesture of an acoustic signal. Background Art
[0002] With the advancement of intelligence, smart cards have become an indispensable part of our lives. Authentication methods based on swiping cards, such as payment and unlocking, can be seen everywhere. While simplifying identity authentication and enhancing user experience, smart cards also face huge security threats. The loss of smart cards, the leakage of card information, and unauthorized copying can all be exploited by attackers, causing huge losses to users. According to Nielsen's annual report, credit card fraud will cause global losses of approximately US$408 billion in the next 10 years. Therefore, there is an urgent need for a tool that can improve the security of card swiping behavior.
[0003] Fortunately, various card swiping behaviors are often accompanied by the action of raising the arm. The arm raising action is different for different individuals, and the action is difficult to imitate or copy. Therefore, choosing the arm raising action as the second factor of identity authentication can effectively improve the security of card swiping behavior. Existing recognition of arm raising actions mainly includes recognition methods based on wearable devices and cameras. Identification based on wearable devices faces unfavorable factors such as the additional cost of the second device, new threats caused by the theft of authentication equipment, and low concealment; camera-based authentication is often easily affected by environmental factors such as light brightness and obstructed vision, so there are also great security risks. Therefore, there is an urgent need for an identification method that is closer to real-life scenarios, easier to promote, and more stable to ensure user information security. Summary of the invention
[0004] In view of the shortcomings of the existing technology, a two-factor identity authentication method based on acoustic signals and arm-raising gesture is provided, which is more concealed, easier to promote and safer.
[0005] To achieve the above technical objectives, the present invention discloses a dual-factor identity authentication method based on acoustic signals and arm-raising gestures. While identifying a handheld smart card through signals, an ultrasonic transceiver is used to identify the arm-raising gesture when swiping the card, thereby realizing dual identification of the smart card and the arm-raising gesture identity.
[0006] The identification of the arm raising gesture includes the following steps:
[0007] Step 1: Use an ultrasonic transceiver device set near the smart card reader to continuously transmit ultrasonic signals to the arm of the card swiping while swiping the card, and continuously receive feedback signals until all ultrasonic feedback signals in a complete card swiping and arm raising gesture are obtained. In this process, the signal reflected back by the arm is reflected back through 16 paths, so the timing data of 16 channels are received;
[0008] Step 2: Calculate the energy spectral density ESD information of the arm-raising action, pre-process the received ultrasonic reflection signal, and the received data of the 16 channels are as follows: the acoustic tone T(T)=sin(2πft+b) with a frequency of f and an initial phase of φ, where t represents the time, f represents the frequency of 48KHz, b represents the bias constant, and T represents the time of the entire arm-raising gesture. The data of the 16 channels are finally received at time T, which is expressed as:
[0009]
[0010] Where Ω = 16 represents the set of all sound signal channels, sampling frequency f = 48KHz, A i is the amplitude coefficient of the acoustic signal of path i. The signals of 16 channels correspond to a 16-dimensional matrix. The signals of 16 channels are summed to obtain a 1×N matrix, and a set of ultrasonic data arranged in time is obtained. The ultrasonic data is detected and replaced with outliers and trend items, and Butterworth filtering and wavelet denoising are used to jointly filter the ultrasonic data to make the ultrasonic data signal more stable and increase the signal-to-noise ratio.
[0011] Step 3: Extract the signal representing the arm-raising action in the ultrasonic data: Select the periodic Hamming window function to divide the ultrasonic data into frames, calculate the power energy spectrum of the segmented signal, use the Hamming window function threshold method to distinguish between action and noise, and use the power spectrum density ESD value by setting the empirical threshold to determine which part is the action and which part is just noise;
[0012] Step 4: Use the collected arm-raising gesture data as a training set to train the improved LSTM-FCN model, so that the improved LSTM-FCN model that has completed the training becomes a model that can exclusively judge and train the arm-raising action;
[0013] Step 5: Input the arm-raising gesture data with judgment into the trained improved LSTM-FCN model. The trained improved LSTM-FCN model extracts the similarity ratio with the user's identity features from the input information center, and then uses a small sample to train the support vector machine (SVM) for classification tasks. By judging the degree of similarity, the arm-raising gesture identity authentication based on the acoustic signal is completed.
[0014] Furthermore, when detecting and replacing outliers in step 2, a sliding window of N=2k+1 needs to be set, where N is the window width, K is the sliding step length, and a threshold A is set to distinguish outliers from normal values; when any sample point in the window differs from the median in the window by A standard deviations, the sample point is judged to be an outlier, and the median is used to replace the outlier; set K=3, A=3; for the detrending term, each data point on the energy spectral density ESD information of the arm-lifting action is expressed as: Y(t)=a+bt+e(t), a represents the intercept of the data, b represents the slope of the data, and e(t) represents the random error term in the data;
[0015] The arm-raising gesture signal is jointly filtered using high-pass filtering and wavelet denoising. The high-pass filter order is 2nd order and the cutoff frequency is 15Hz. For wavelet denoising, the wavelet coefficients exceeding the threshold are considered to be generated by the arm-raising gesture signal and are retained. On the contrary, the coefficients below the threshold are considered to be generated by noise and are set to zero for denoising. The threshold is determined according to the signal-to-noise ratio of the original signal.
[0016] Furthermore, the improved LSTM-FCN model includes a long short-term memory network LSTM and a fully convolutional neural network FCN connected in parallel with the input. The output of the long short-term memory network LSTM and the output mean of the fully convolutional neural network FCN are mixed through a global pooling Concat function with a channel number of 256 and then output to the support vector machine SVM; the long short-term memory network LSTM includes a series of LSTM-1 and LSTM-2, and LSTM-1 and LSTM-2 contain 64 and 128 neuron inputs and outputs respectively; the fully convolutional neural network FCN includes a series of Conv-1 and Conv-2, Conv-1 is a series connection of three groups of Conv, the number of channels is 64, and the number of channels of Conv-2 is 128; Conv-1 is a 3×3 convolution block, which extracts effective feature information of the signal through the neighborhood induction of the convolution. After the 3×3 convolution, a point-by-point convolution with a convolution kernel of 1×1 is connected. After each convolution layer, a compression and excitation module (Squeeze-and-Excitation) is set. The compression and excitation module includes two sub-modules, the compression module and the excitation module. After the compression and excitation module, a batch normalization layer (BatchNorm) is set, and the activation function GELU is connected after the batch normalization layer;
[0017] There is a Dropout layer after LSTM-1 and LSTM-2. The probability of the Dropout layer is set to 0.5, which means that during the training process, each neuron has a 50% probability of being randomly dropped.
[0018] During training, the concatenated vector of the FCN and LSTM outputs is passed through a fully connected layer and then input into the Softmax function for classification prediction. The Softmax function calculates the probability distribution of each category to determine the possibility that the input data belongs to each category.
[0019] When training the model, the data is labeled according to the determined positive sample category, and the time series information and convolution features of the collected ultrasonic gesture signal are extracted through the LSTM and FCN networks respectively, and the feature vector is passed as input to a layer of SVM classifier. SVM divides data points into different categories by constructing an optimal hyperplane in the feature space. The model trains the LSTM-FCN feature extractor by extracting time series signal features that are not related to the category, and then uses a small sample to train the support vector machine (SVM) for classification tasks.
[0020] Furthermore, the setting of the ultrasonic transceiver can be selected according to needs, including: the ultrasonic transceiver is set near the card swiping receiver to ensure that the recognition process is not blocked; using a mobile phone as an ultrasonic transceiver, it is only necessary to ensure that the mobile phone is placed near the arm raising action, using the mobile phone speaker to emit sound waves, and using the microphone to receive them.
[0021] An ultrasonic arm-raising action authentication device comprises: an ultrasonic generating device, an ultrasonic receiving device, a processor, and a memory; the ultrasonic generating device is used to emit ultrasonic waves in response to an arm-raising gesture made by a nearby person who swipes a card; the ultrasonic receiving device is used to receive a feedback signal of the emitted ultrasonic wave; the memory is used to store program code; the processor is used to call the program code; so as to execute the above-mentioned arm-raising gesture two-factor identity authentication method based on acoustic signals.
[0022] A computer-readable storage medium stores a computer program, which is suitable for being loaded by a processor and executing the above-mentioned arm-raising gesture two-factor identity authentication method based on acoustic signals.
[0023] Beneficial effects: The present invention proposes a dual-factor identity authentication method based on acoustic signals and designs an innovative signal detection algorithm for arm-raising behavior. It can effectively extract the arm-raising gesture signal on the basis of card swiping recognition, and use the recognition of the arm-raising gesture to determine whether the card swipeer is the correct holder of the identity card, thereby achieving double insurance on the basis of card recognition. The present invention extracts user features through an improved LSTM-FCN network, and then constructs an authentication model in combination with an SVM classifier. The pre-trained model is extended to different arm-raising gesture recognition tasks through fine-tuning transfer learning. This method has achieved significant improvements in both accuracy and robustness. It effectively solves the security risks existing in existing identity authentication, and improves the accuracy and stability of authentication by optimizing the model. BRIEF DESCRIPTION OF THE DRAWINGS
[0024] Figure 1 Schematic diagram of obtaining an arm-raising gesture acoustic signal in an embodiment of the present invention.
[0025] Figure 2 It is an overall waveform diagram comparing the effects before and after signal noise reduction in an embodiment of the present invention.
[0026] Figure 3 It is a performance bar chart of the detection accuracy of the method of the present invention after experimental comparison.
[0027] Figure 4 It is the working characteristic ROC curve of all user identity recognition in the embodiment of the present invention.
[0028] Figure 5 It is a performance comparison chart of the user authentication model constructed in the embodiment of the present invention and the existing network.
[0029] Figure 6 It is a bar chart showing the security evaluation of the present invention when it is subjected to three attack situations in the experiment. DETAILED DESCRIPTION
[0030] The embodiments of the present invention are further described with reference to the accompanying drawings:
[0031] The present invention discloses a double-factor identity authentication method based on an arm-raising gesture of an acoustic signal. An ultrasonic device is arranged near the card swiping location to collect and identify the cardholder's actions during the card swiping process, thereby reconfirming whether the cardholder is the card holder. The arm-raising action during the card swiping process is detected by the following process:
[0032] Step 1: Data acquisition: Acquire the acoustic signal reflected by the user's arm-raising action.
[0033] Step 2: Signal detection and extraction. First, the channels of the 16 channels are summed up, and then the superimposed channel data is jointly denoised. In the algorithm, a 1×N matrix is obtained, and outliers, trend items, and Butterworth filtering and wavelet denoising are selected for joint filtering. After completing the above work, the power energy spectrum of the segmented signal is calculated, and the window function threshold method is used to distinguish between action and noise, and the action is extracted.
[0034] Step 3: Improve the LSTM-FCN model by combining the Squeeze-and-Excitation module and the activation function GELU. Use the model network structure to extract the deep features of the signal, and then use small sample training support vector machine (SVM) to perform classification tasks to achieve user identity authentication.
[0035] The present invention mainly solves the problems existing in the existing identity authentication methods; it provides a two-factor identification method that is more concealed, closer to real-life scenarios, and easier to promote for card swiping authentication scenarios, and solves the problems existing in existing identity identification, such as being easily affected by environmental factors and requiring additional equipment. The method first uses the user's smart device to emit ultrasonic waves, and then deploys it on a smart phone in the form of an Android APP to obtain the acoustic signal reflected by the user's arm-raising action. The obtained acoustic signal is pre-processed to extract the arm-raising action signal. A high-performance and lightweight identity authentication model is constructed, and then the pre-trained model is applied to the arm-raising gesture recognition task of the acoustic signal through transfer learning.
[0036] like Figure 1 As shown, the specific method flow of the present invention is as follows:
[0037] Input: arm-raising gesture.
[0038] Output: Identification result
[0039] 1. In the data acquisition stage, the energy spectral density (ESD) of the acoustic signal describes the distribution of the energy of the acoustic signal in space with frequency, and can sense the movement of all surrounding objects as the energy distribution changes. Therefore, the present invention uses the ESD of the acoustic signal obtained by the smartphone. It can be deployed on the mobile phone as an Android APP. The built-in speaker of the smartphone emits ultrasonic waves, and then the microphone receives the acoustic signal reflected by the arm raising action.
[0040] 2. In the signal detection and extraction stage, the 16 channels obtained are first summed in the algorithm to obtain a 1×N matrix, and the signal segments with large noise amplitudes at the beginning and end are removed. The hampel filtering method is used to detect and delete outliers from the collected time series data. The linear trend component in the data is obtained by finding the best fitting straight line of the data, and it is subtracted from the data to achieve the effect of removing the trend term. Suppose there is a set of time series data Y(t) with a length of N, where t represents the time step. For linear trend removal, each data point can be expressed as:
[0041] Y(t)=a+bt+e(t)
[0042] The solution for the best fit line is found by minimizing the sum of squared errors:
[0043]
[0044] Subtract this linear trend component from the original data to obtain the detrended time series data. After completing the above processing, the data also includes interference signals generated by environmental noise and human body jitter. The present invention uses a high-pass filter and wavelet denoising method to smooth the signal. The formula for generating the Butterworth high-pass filter is:
[0045]
[0046] Where n is the order of the filter, and D0 is the cutoff frequency of the Butterworth high-pass filter. When using wavelet denoising, assume that the ideal original signal is
[0047] d(t)=f(t)+σz(t)
[0048] Among them, f(t) is the signal not contaminated by noise, z(t) is the noise, and σ is the noise intensity. The signal d(t) is subjected to wavelet decomposition to obtain a set of wavelet decomposition coefficients Wj. The wavelet coefficients Wj are subjected to threshold processing, and then the signal is reconstructed by inverse wavelet transform to obtain the output signal f'(t), which is the signal after noise reduction. Joint filtering aims to minimize the interference caused by noise while retaining important action information, thereby facilitating signal segmentation. After completing the preprocessing of the data, the present invention selects a periodic hamming window for framing, calculates the power energy spectrum of the segmented signal respectively, and judges the location where the action occurs based on the difference in power spectrum density, such as Figure 2 Shown are schematic diagrams before and after noise reduction.
[0049] In the stage of building the identity authentication model, based on the information obtained and combining the advantages of deep learning and traditional machine learning, a paradigm model based on pre-training and fine-tuning was established. By using the improved LSTM-FCN feature extractor, a small sample was used to train the support vector machine (SVM) for classification tasks.
[0050] During the feature extraction period, an improved LSTM-FCN network was used. In FCN, two convolution blocks were used for stacking. The first block consisted of output channels with 64 features, and the second block had 128 output channels. The initial block was a 3×3 convolution. In order to optimize the model performance and fusion characteristics, a 1×1 point-by-point convolution was used on the original basis of the network to reduce the number of parameters. After learning the weights of different convolution kernels, the network adjusted and merged the features from different channels to enhance the representation ability of the model. In addition, a squeeze and excitation module was added to each convolution layer. In the squeeze module, a global average pooling operation was used to compress the feature map of each channel to capture the global sensory field of the channel. In the excitation module, a small fully connected network including two fully connected layers was used. The feature vector obtained in the previous step was used as input to learn the excitation weights of the channel. These weights can adaptively adjust the importance of each channel to enhance the representation of different features. After the excitation, GELU was selected as the activation function of the post-batch norm layer to normalize and standardize the input, which accelerated the training process of the model. In addition, two LSTM blocks are adopted and combined with Dropout layers to effectively learn and capture the temporal information in the input sequence. During user authentication and transfer learning, fine-tuning-based transfer learning is used to transfer previously acquired knowledge and models to new learning tasks. The LSTM-FCN feature extractor is first used to extract category-independent temporal signal features, and then a small sample radial basis function kernel is used to train the SVM for authentication, which is finally used to solve the new learning task.
[0051] like Figure 2 As shown in the figure, the experimental results show that the proposed data processing method can effectively reduce the noise in the signal. Figure 3 As shown in the figure, the experimental results show that the average false negative rate (FNR) of the proposed arm-raising gesture authentication method based on acoustic signals is 4.973%, the average false positive rate (FPR) is 5.936%, and the average detection accuracy is 97.191%, which has high accuracy. Figure 4 , Figure 5 It can be seen that the proposed authentication model can show good performance in both user identification and authentication. Compared with the existing LSTM-FCN network, the proposed model improves the authentication accuracy by 2.4%, reduces the FAR and FRR values, and significantly reduces the number of parameters. Figure 6As shown, the detection accuracy of brute force attacks is 97.414%, the detection accuracy of imitation attacks is 97.170%, and the detection accuracy of replay attacks is 97.273%, which shows that it has high security.
[0052] Therefore, it has the following advantages: the detection behavior is more covert, the detected action is close to the real life scene, no additional equipment is required, and it is easier to promote. The LSTM-FCN network is used to build a deep learning model for pre-training, and then the SVM classifier is used for classification, which improves the accuracy and robustness, so that the proposed identity authentication method has higher security.
Claims
1. A two-factor identity authentication method based on acoustic signals and arm-raising gestures. It is characterized in that While identifying the handheld smart card through signals, the ultrasonic transceiver is used to identify the arm-raising gesture when swiping the card, thereby realizing dual identification of the smart card and the arm-raising gesture identity; The identification of the arm raising gesture includes the following steps: Step 1: Use an ultrasonic transceiver device set near the smart card reader to continuously transmit ultrasonic signals to the arm of the card swiping while swiping the card, and continuously receive feedback signals until all ultrasonic feedback signals in a complete card swiping and arm raising gesture are obtained. Assume that the signal reflected back by the arm in this process is reflected back through 16 paths, so the timing data of 16 channels are received; Step 2: Calculate the energy spectral density ESD information of the arm-raising action, pre-process the received ultrasonic reflection signal, and the received data of the 16 channels are as follows: the acoustic tone T(T)=sin(2πft+b) with a frequency of f and an initial phase of φ, where t represents the time, f represents the frequency of 48KHz, b represents the bias constant, and T represents the time of the entire arm-raising gesture. The data of the 16 channels are finally received at time T, which is expressed as: Where Ω = 16 represents the set of all sound signal channels, sampling frequency f = 48KHz, A i is the amplitude coefficient of the acoustic signal of path i. The signals of 16 channels correspond to a 16-dimensional matrix. The signals of 16 channels are summed to obtain a 1×N matrix, and a set of ultrasonic data arranged in time is obtained. The ultrasonic data is detected and replaced with outliers and trend items, and Butterworth filtering and wavelet denoising are used to jointly filter the ultrasonic data to make the ultrasonic data signal more stable and increase the signal-to-noise ratio. Step 3: Extract the signal representing the arm-raising action in the ultrasonic data: Select the periodic Hamming window function to divide the ultrasonic data into frames, calculate the power energy spectrum of the segmented signal, use the Hamming window function threshold method to distinguish between action and noise, and use the power spectrum density ESD value by setting the empirical threshold to determine which part is the action and which part is just noise; Step 4: Use the collected arm-raising gesture data as a training set to train the improved LSTM-FCN model, so that the improved LSTM-FCN model that has completed the training becomes a model that can exclusively judge and train the arm-raising action; Step 5: Input the arm-raising gesture data with judgment into the trained improved LSTM-FCN model. The trained improved LSTM-FCN model extracts the similarity ratio with the user's identity features from the input information center, and then uses a small sample to train the support vector machine (SVM) for classification tasks. By judging the degree of similarity, the arm-raising gesture identity authentication based on the acoustic signal is completed.
2. According to claim 1, a two-factor identity authentication method based on acoustic signal and arm-raising gesture, Features: When detecting and replacing outliers in step 2, a sliding window of N=2k+1 needs to be set, where N is the window width, K is the sliding step length, and a threshold A is set to distinguish outliers from normal values; when any sample point in the window differs from the median in the window by A standard deviations, the sample point is judged to be an outlier, and the median is used to replace the outlier; set K=3, A=3; for the detrending term, each data point on the energy spectral density ESD information of the arm lifting action is expressed as: Y(t)=a+bt+e(t), a represents the intercept of the data, b represents the slope of the data, and e(t) represents the random error term in the data; The arm-raising gesture signal is jointly filtered using high-pass filtering and wavelet denoising. The high-pass filter order is 2nd order and the cutoff frequency is 15Hz. For wavelet denoising, the wavelet coefficients exceeding the threshold are considered to be generated by the arm-raising gesture signal and are retained. On the contrary, the coefficients below the threshold are considered to be generated by noise and are set to zero for denoising. The threshold is determined according to the signal-to-noise ratio of the original signal.
3. According to claim 1, a double-factor identity authentication method based on acoustic signal and arm-raising gesture, Features: The improved LSTM-FCN model includes a long short-term memory network LSTM and a fully convolutional neural network FCN connected in parallel with the input. The output of the long short-term memory network LSTM and the output mean of the fully convolutional neural network FCN are mixed through a global pooling Concat function with a channel number of 256 and then output to the support vector machine SVM; the long short-term memory network LSTM includes a series of LSTM-1 and LSTM-2, and LSTM-1 and LSTM-2 contain 64 and 128 neurons in input and output respectively; the fully convolutional neural network FCN includes a series of Conv -1 and Conv-2, Conv-1 is three groups of Conv connected in series, the number of channels is 64, and the number of channels of Conv-2 is 128; Conv-1 is a 3×3 convolution block, which extracts effective feature information of the signal through the neighborhood induction of the convolution. After the 3×3 convolution, a point-by-point convolution with a convolution kernel of 1×1 is connected. After each convolution layer, a compression and excitation module (Squeeze-and-Excitation) is set. The compression and excitation module includes two sub-modules, the compression module and the excitation module. After the compression and excitation module, a batch normalization layer (Batch Norm) is set, and the activation function GELU is connected after the batch normalization layer; There is a Dropout layer after LSTM-1 and LSTM-2. The probability of the Dropout layer is set to 0.5, which means that during the training process, each neuron has a 50% probability of being randomly dropped. During training, the concatenated vector of the FCN and LSTM outputs is passed through a fully connected layer and then input into the Softmax function for classification prediction. The Softmax function calculates the probability distribution of each category to determine the possibility that the input data belongs to each category. When training the model, the data is labeled according to the determined positive sample category, and the time series information and convolution features of the collected ultrasonic gesture signal are extracted through the LSTM and FCN networks respectively, and the feature vector is passed as input to a layer of SVM classifier. SVM divides data points into different categories by constructing an optimal hyperplane in the feature space. The model trains the LSTM-FCN feature extractor by extracting time series signal features that are not related to the category, and then uses a small sample to train the support vector machine (SVM) for classification tasks.
4. According to claim 1, a double-factor identity authentication method based on acoustic signal and arm-raising gesture, Features: The setting precautions of the ultrasonic transceiver are selected according to needs, including: the ultrasonic transceiver is set near the card swiping receiver to ensure that the recognition process is not blocked; using a mobile phone as an ultrasonic transceiver, just ensure that the mobile phone is placed near the arm raising action, use the mobile phone speaker to emit sound waves, and use the microphone to receive them.
5. An ultrasonic arm-raising action authentication device, It is characterized in that include: Ultrasonic wave generating device, ultrasonic wave receiving device, processor, and memory; The ultrasonic generating device is used to emit ultrasonic waves when a nearby person swipes a card or raises an arm; the ultrasonic receiving device is used to receive feedback signals of the emitted ultrasonic waves; the memory is used to store program codes; and the processor is used to call the program codes; To execute the double-factor identity authentication method based on acoustic signal and arm-raising gesture as described in any one of claims 1-3.
6. A computer-readable storage medium, It is characterized in that The computer-readable storage medium stores a computer program, which is suitable for being loaded by a processor and executing the two-factor identity authentication method based on acoustic signals and arm-raising gestures as described in any one of claims 1 to 3.
Citation Information
Patent Citations
Gesture identification method and system based on mobile terminal
CN103226386A
Ultrasonic wave based volley gesture identification method and system
CN105807923A
Ultrasonic gesture recognition method and system
CN106203380A
Gesture recognition method and system based on dual-track audio Doppler effect
CN107450724A
Ultrasonic processing method and device, electronic equipment and computer readable medium
CN110764612A
