A speech recognition method, system, device, and medium for elevator entrapment scenarios.

Abstract

This invention provides a speech recognition method, system, device, and medium for elevator entrapment scenarios. The method includes: acquiring speech data in an elevator scenario and preprocessing the speech data to obtain a first speech feature; using the first speech feature as input to a deep neural network to output recognized text and an entrapment probability value; determining entrapment based on the recognized text and the entrapment probability value, and outputting an entrapment determination result. By using a deep neural network and a specific encoding and decoding process, this method can more accurately process speech data in elevator scenarios, thereby improving the accuracy of entrapment detection. By comparing the entrapment probability value with a threshold and combining it with the recognized text for comprehensive judgment, this method can reduce false detections and false negatives, improving the accuracy of entrapment determination.

Description

Technical Field

[0001] This invention relates to the field of elevator safety technology, specifically to a voice recognition method, system, device, and medium for elevator entrapment scenarios. Background Technology

[0002] With the rapid development of deep learning technology in recent years, speech recognition technology has been widely applied in various industries, bringing convenience and innovation to people's lives. However, in certain specific scenarios, such as detecting whether someone is trapped in an elevator using speech recognition, speech recognition models may face some challenges and limitations. In such scenarios, existing technologies often only convert speech into text and judge the entrapment event by matching keywords, which often fails to meet actual needs. Moreover, due to factors such as noise in the speech signal, accent differences, and speaking environment, the accuracy of speech recognition is not high enough, leading to false positives and false negatives. These false positives and false negatives may adversely affect user experience and safety, thus requiring more accurate and reliable methods for detecting entrapment events. Summary of the Invention

[0003] In view of the shortcomings of the prior art described above, the present invention provides a voice recognition method, system, device and medium for elevator entrapment scenarios to solve the above-mentioned technical problems.

[0004] This invention provides a speech recognition method for elevator entrapment scenarios, comprising:

[0005] Acquire voice data in an elevator scenario and preprocess the voice data to obtain a first voice feature;

[0006] The first speech feature is used as the input to a deep neural network, and the output is the recognized text and the probability value of trapping a person.

[0007] Based on the identified text and the probability value of being trapped, a trapping determination is made, and the trapping determination result is output.

[0008] In this invention, the preprocessing includes:

[0009] The duration of the voice data is obtained, and the voice data is segmented according to a preset time interval to obtain segmentation features;

[0010] The frequency of the speech data is obtained, and the feature length of the segmented feature is calculated based on the time interval and the frequency.

[0011] The first speech feature is obtained by combining the feature length of the segmented feature and the number of segments.

[0012] In this invention, the deep neural network is a convolutional neural network.

[0013] In this invention, the step of using the first speech feature as input to a deep neural network and outputting the recognized text and the probability value of trapping the person specifically includes:

[0014] The first speech feature is used as input and fed into the first convolutional layer for calculation to obtain the second speech feature.

[0015] The second speech feature is used as input and fed into a four-layer transformer for encoding to obtain the output code.

[0016] The output encoding is decoded according to a preset dictionary to obtain the recognized text;

[0017] The output code is used as input to the second convolutional layer for calculation to obtain the output code features;

[0018] The output encoded features are used as input and fed into two fully connected layers for calculation to obtain feature values ​​with a dimension of one.

[0019] The feature value is input into a preset trapping probability function to obtain the trapping probability value.

[0020] In this invention, the step of decoding the output encoding according to a preset dictionary to obtain the recognized text specifically includes:

[0021] A probability matrix is ​​generated based on the dictionary, wherein each row vector of the probability matrix contains 3001 probability values, the first 3000 probability values ​​correspond to each Chinese character in the dictionary, and the 3001st probability value corresponds to the empty character;

[0022] Based on the decoding and the probability matrix, the position with the highest probability value is obtained, that is, the character with the highest probability is found, and the output is the recognized character;

[0023] Repeat until all the recognized characters in the encoding are obtained, and perform post-processing on the recognized characters to generate recognized text. The post-processing includes removing duplicate characters among adjacent characters and removing empty characters.

[0024] In this invention, the probability function for trapping a person is:

[0025]

[0026] Where f(x) is the probability function of trapping people, and x is the eigenvalue.

[0027] In this invention, the step of determining whether someone is trapped based on the identified text and the trapping probability value, and outputting the trapping determination result, specifically includes:

[0028] When the probability value of entrapment is less than or equal to the first threshold, the entrapment determination result is that no entrapment event has occurred;

[0029] When the probability value of trapping a person is greater than the first threshold and less than the second threshold, keyword matching is performed on the identified text. When a keyword exists, the trapping result is that a trapping event has occurred. When no keyword exists, the trapping result is that no trapping event has occurred.

[0030] When the probability value of entrapment is greater than the second threshold, the entrapment determination result is that an entrapment event has occurred.

[0031] The present invention also provides a voice recognition system for elevator entrapment scenarios, comprising:

[0032] Data acquisition module: Acquires voice data in the elevator scenario and preprocesses the voice data to obtain the first voice feature;

[0033] Output information module: The first speech feature is used as the input of the deep neural network, and the output is the recognized text and the probability value of trapping people;

[0034] Comprehensive Judgment Module: Based on the identified text and the probability value of being trapped, the module determines whether someone is trapped and outputs the trapping result.

[0035] The present invention also provides an electronic device, the electronic device comprising:

[0036] One or more processors;

[0037] A storage device for storing one or more programs, which, when executed by one or more processors, cause the electronic device to implement a voice recognition method for an elevator entrapment scenario as described above.

[0038] The present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a computer processor, causes the computer to perform a voice recognition method for an elevator entrapment scenario as described in any of the preceding claims.

[0039] This invention provides a speech recognition method, system, device, and medium for elevator entrapment scenarios. The method acquires speech data from an elevator scenario and preprocesses the speech data to obtain a first speech feature. The first speech feature is used as input to a deep neural network, outputting a recognized text and an entrapment probability value. An entrapment determination is made based on the recognized text and the entrapment probability value, and an entrapment determination result is output. The beneficial effects include:

[0040] 1. Improve the accuracy of entrapment event detection: By using deep neural networks and specific encoding and decoding processes, this method can process voice data in elevator scenarios more accurately, thereby improving the accuracy of entrapment event detection.

[0041] 2. Real-time performance and high efficiency: This method uses deep learning technology, which has strong computing power and parallel processing capabilities, enabling real-time speech recognition and trap detection, thus improving response speed and processing efficiency.

[0042] 3. Automation and intelligence: This method uses deep neural networks to automatically encode and decode speech data without human intervention, achieving intelligent recognition of entrapment events and reducing the need for manual operation.

[0043] 4. Reduce false positives and false negatives: By comparing the probability value of entrapment with a threshold and combining it with the recognized text for comprehensive judgment, this method can reduce false positives and false negatives and improve the accuracy of entrapment events.

[0044] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the following are specific embodiments of this application. Attached Figure Description

[0045] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. In the drawings:

[0046] Figure 1 A flowchart illustrating a speech recognition method for an elevator entrapment scenario, as shown in an exemplary embodiment of the present invention;

[0047] Figure 2 A flowchart illustrating the preprocessing in a speech recognition method for an elevator entrapment scenario, as shown in an exemplary embodiment of the present invention;

[0048] Figure 3 A flowchart illustrating the output of a deep neural network in a speech recognition method for an elevator entrapment scenario, as shown in an exemplary embodiment of the present invention;

[0049] Figure 4 This is a schematic diagram of a voice recognition system for an elevator entrapment scenario, which is an exemplary embodiment of the present invention. Detailed Implementation

[0050] The embodiments of the present invention will be described below with reference to the accompanying drawings and preferred embodiments. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be understood that the preferred embodiments are only for illustrating the present invention and not for limiting the scope of protection of the present invention.

[0051] It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of the present invention. Therefore, the drawings only show the components related to the present invention and are not drawn according to the actual number, shape and size of the components in the actual implementation. In the actual implementation, the form, quantity and proportion of each component can be arbitrarily changed, and the layout of the components may also be more complex.

[0052] In the following description, numerous details are explored to provide a more thorough explanation of embodiments of the invention. However, it will be apparent to those skilled in the art that embodiments of the invention may be practiced without these specific details. In other embodiments, well-known structures and devices are shown in block diagram form rather than in detail to avoid obscuring embodiments of the invention.

[0053] First, it's important to clarify that a Convolutional Neural Network (CNN) is a deep learning algorithm specifically designed to process data with a grid-like structure, such as images and speech. It has achieved tremendous success in computer vision and performs exceptionally well in many tasks, including image recognition, object detection, and image generation.

[0054] The main characteristics of CNNs are their local connectivity and weight-sharing mechanisms. They utilize convolutional layers to perform convolution operations on the input data, extracting features of different sizes. The convolution operation involves sliding a filter (also called a convolutional kernel) across the input data, multiplying the filter weights by the corresponding regions of the input data, and summing the results to generate an output feature map.

[0055] During convolution, the weight-sharing feature allows inputs from different regions to share the same weights, which greatly reduces the number of parameters that need to be learned, thereby reducing the complexity of the network, reducing the risk of overfitting, and speeding up the training of the model.

[0056] CNNs typically consist of multiple convolutional layers, activation function layers (such as ReLU layers), pooling layers (such as MaxPooling layers), and fully connected layers. Pooling layers are used to reduce the size of feature maps, decrease computational complexity, and extract key features. Fully connected layers connect the outputs of the pooling layers to the output layer, and are used for tasks such as classification and regression.

[0057] CNNs are trained using the backpropagation algorithm, optimizing weights to minimize the loss function so that the network output is as close as possible to the true label. After training on large-scale datasets, CNNs can learn effective feature representations and make accurate predictions on unseen data.

[0058] Figure 1 A flowchart illustrating a speech recognition method for an elevator entrapment scenario, as shown in an exemplary embodiment of the present invention;

[0059] like Figure 1 As shown, the present invention provides a speech recognition method for elevator entrapment scenarios, comprising:

[0060] Step S110: Acquire voice data in the elevator scenario and preprocess the voice data to obtain the first voice feature;

[0061] Step S120: Use the first speech feature as input to the deep neural network, and output the recognized text and the probability value of trapping people;

[0062] Step S130: Determine if someone is trapped based on the identified text and the trapped probability value, and output the trapped determination result.

[0063] Specifically, in step S110, the acquired speech data is preprocessed to obtain the first speech feature. The purpose is to extract key features from the speech data, providing more useful information for subsequent speech recognition. In step S120, the powerful learning ability of deep neural networks enables them to learn complex speech feature representations from large amounts of data, thereby improving the accuracy and robustness of speech recognition. In step S130, by comprehensively considering the recognized text and the probability value of entrapment, the accuracy of judging entrapment events can be improved, and entrapment events within different probability ranges can be distinguished, increasing the system's flexibility and adjustability.

[0064] Figure 2 A flowchart illustrating the preprocessing in a speech recognition method for an elevator entrapment scenario, as shown in an exemplary embodiment of the present invention;

[0065] like Figure 2 As shown, the present invention provides a speech recognition method for elevator entrapment scenarios, wherein the preprocessing includes:

[0066] Step S210: Obtain the duration of the voice data, segment the voice data according to a preset time interval, and obtain segmentation features;

[0067] Step S220: Obtain the frequency of the speech data, and calculate the feature length of the segmented feature based on the time interval and the frequency;

[0068] Step S230: Based on the feature length of the segmented features and the number of segments, the first speech feature is obtained by combining them.

[0069] Specifically, in step S210, the duration T of the speech data is obtained. Based on a preset time interval t, the speech data is segmented into N = [T / t] segments, where [] is a rounding function and rounds up. The duration of each speech segment is the preset time interval t, and the last segment is padded with space if its duration is insufficient. In step S220, the frequency hHz of the speech data is obtained, and the feature length of each segment is L = t*h. In step S230, based on the feature length L of the segmented feature and the number of segments N, a first speech feature of size N*L is finally obtained.

[0070] In one exemplary embodiment, the deep neural network is a convolutional neural network.

[0071] Specifically, when training the convolutional neural network, historical voice samples in elevator scenarios are collected and divided into two categories: key voice samples for entrapment, with n samples; and ordinary voice samples, which include all collected samples excluding key voice samples for entrapment, with m samples.

[0072] For each audio sample, the following steps are taken: label the text corresponding to the audio; label whether it is a key audio element that is tampering with people. If it is a key audio element that is tampering with people, label it with 1; otherwise, label it with 0.

[0073] Based on the constructed elevator entrapment speech recognition model, 24 rounds of model training were conducted using samples, corresponding text labels, and corresponding entrapment key labels. The elevator entrapment speech recognition model includes a speech recognition module, an entrapment probability module, and an entrapment judgment module. The speech recognition module, based on an encoder-decoder structure, converts speech into text; the entrapment probability module outputs the probability that the current speech is related to an entrapment event; and the entrapment judgment module combines the text output from the previous two modules with the entrapment probability to determine whether the elevator is trapping someone.

[0074] Figure 3 A flowchart illustrating the output of a deep neural network in a speech recognition method for an elevator entrapment scenario, as shown in an exemplary embodiment of the present invention;

[0075] like Figure 3As shown, the present invention provides a speech recognition method for elevator entrapment scenarios, wherein the first speech feature is used as input to a deep neural network, and the output is the recognized text and the entrapment probability value, specifically including:

[0076] Step S310: The first speech feature is used as input and fed into the first convolutional layer for calculation to obtain the second speech feature;

[0077] Step S320: The second speech feature is used as input and fed into a four-layer transformer for encoding to obtain the output encoding;

[0078] Step S330: Decode the output encoding according to a preset dictionary to obtain the recognized text;

[0079] Step S340: The output code is used as input and fed into the second convolutional layer for calculation to obtain the output code features;

[0080] Step S350: The output encoded features are used as input and fed into two fully connected layers for calculation to obtain feature values ​​with a dimension of one;

[0081] Step S360: Input the feature value into a preset trapping probability function to obtain the trapping probability value.

[0082] Specifically, in step S310, the first speech feature of size N*L from step 230 is convolved twice with a 3*3 convolution and a stride of 2 through two convolutional layers to obtain a second speech feature of size N*(L / 4). In step S320, the second speech feature is encoded through four transformer layers to obtain an output encoding of size N*256. In step S330, the output encoding is decoded according to a preset dictionary containing 3000 commonly used Chinese characters to obtain the recognized text. In steps S340, S350, and S360, the output encoding is used as input; the encoding is convolved twice with a size of 3*3 and a stride of 2 to obtain an output encoding feature of size N*64; the output encoding feature is input to two fully connected layers to obtain a feature value of dimension one; the feature value is input to a preset trapping probability function to obtain a trapping probability value p.

[0083] In one exemplary embodiment, the step of decoding the output encoding according to a preset dictionary to obtain the recognized text specifically includes:

[0084] A probability matrix is ​​generated based on the dictionary, wherein each row vector of the probability matrix contains 3001 probability values, the first 3000 probability values ​​correspond to each Chinese character in the dictionary, and the 3001st probability value corresponds to the empty character;

[0085] Based on the decoding and the probability matrix, the position with the highest probability value is obtained, that is, the character with the highest probability is found, and the output is the recognized character;

[0086] Repeat until all the recognized characters in the encoding are obtained, and perform post-processing on the recognized characters to generate recognized text. The post-processing includes removing duplicate characters among adjacent characters and removing empty characters.

[0087] Specifically, a probability matrix is ​​generated based on a dictionary containing 3000 commonly used Chinese characters. Each row of the vector contains 3001 probability values. The first 3000 probabilities correspond to the total number of characters in the dictionary, and the 3001st probability value corresponds to an empty character. The character in the dictionary corresponding to the highest probability value in each row is the output. A total of N characters are obtained. First, duplicate characters in adjacent characters are eliminated, and then empty characters are removed to finally obtain the recognized text, which is in TXT format.

[0088] In one exemplary embodiment, the probability function for trapping a person is:

[0089]

[0090] Where f(x) is the probability function of trapping people, and x is the eigenvalue.

[0091] Specifically, by substituting the feature values ​​into the trap probability function f(x), we can guarantee that the output value range is (0,1).

[0092] In one exemplary embodiment, the step of determining whether someone is trapped based on the identified text and the trapping probability value, and outputting the trapping determination result, specifically includes:

[0093] When the probability value of entrapment is less than or equal to the first threshold, the entrapment determination result is that no entrapment event has occurred;

[0094] When the probability value of trapping a person is greater than the first threshold and less than the second threshold, keyword matching is performed on the identified text. When a keyword exists, the trapping result is that a trapping event has occurred. When no keyword exists, the trapping result is that no trapping event has occurred.

[0095] When the probability value of entrapment is greater than the second threshold, the entrapment determination result is that an entrapment event has occurred.

[0096] Specifically, in a feasible embodiment, the first threshold is set to 0.35 and the second threshold is set to 0.65. Then: if p <= 0.35, it is directly determined that no entrapment event occurs; if 0.35 < p < 0.65, keywords such as "cannot open", "door malfunction", "help" are matched in the text. If the match is successful, it is determined that there is an entrapment event, otherwise there is no entrapment event; if p >= 0.65, it is directly determined that there is an entrapment event. By jointly identifying the text and the probability p value for judgment, many false detections can be filtered out and missed detections can be avoided.

[0097] The comparative test results of a voice recognition method for elevator entrapment scenarios provided by this method and single voice recognition + keyword matching are shown in the following table. Among them, recall rate = the number of detected real entrapment events / the total number of all real entrapment events. The larger this value is, the fewer missed detections; precision = the number of detected real entrapment events / the number of detected entrapment events. The larger this value is, the fewer false detections.

[0098] Recall rate accuracy Speech recognition + keyword matching 93.6％ 95.1％ A speech recognition method for elevator entrapment scenarios 96.1％ 97.4％

[0099] Figure 4 It is a schematic structural diagram of a voice recognition system for elevator entrapment scenarios shown in an exemplary embodiment of the present invention;

[0100] As Figure 4 shown, the exemplary voice recognition system for elevator entrapment scenarios includes:

[0101] Data acquisition module 401: Acquire voice data in the elevator scenario and preprocess the voice data to obtain the first voice feature;

[0102] Output information module 402: Use the first voice feature as the input of the deep neural network and output the recognized text and entrapment probability value;

[0103] Comprehensive determination module 403: Make an entrapment determination based on the recognized text and entrapment probability value and output the entrapment determination result.

[0104] It should be noted that the voice recognition system for elevator entrapment scenarios provided in the above embodiment and the voice recognition method for elevator entrapment scenarios provided in the above embodiment belong to the same concept. The specific ways in which each module and unit perform operations have been described in detail in the method embodiment and will not be repeated here. In actual application, the voice recognition system for elevator entrapment scenarios provided in the above embodiment can, as needed, allocate the above functions to different functional modules, that is, divide the internal structure of the system into different functional modules to complete all or part of the functions described above. This is not limited here either.

[0105] Embodiments of this application also provide an electronic device, including: one or more processors; and a storage device for storing one or more programs, which, when executed by the one or more processors, enable the electronic device to implement a voice recognition method for an elevator entrapment scenario provided in the above embodiments.

[0106] Embodiments of this application also provide a computer system for an electronic device. It should be noted that the computer system for the electronic device is merely an example and should not impose any limitations on the functionality or scope of use of the embodiments of this application.

[0107] Specifically, the computer system includes a Central Processing Unit (CPU), which can perform various appropriate actions and processes based on programs stored in Read-Only Memory (ROM) or loaded from storage into Random Access Memory (RAM), such as executing the methods described in the above embodiments. The RAM also stores various programs and data required for system operation. The CPU, ROM, and RAM are interconnected via a bus. Input / output (I / O) interfaces are also connected to the bus.

[0108] The following components are connected to the I / O interface: input sections including keyboards, mice, etc.; output sections including cathode ray tubes (CRTs), liquid crystal displays (LCDs), and speakers; storage sections including hard drives; and communication sections including network interface cards such as LAN (Local Area Network) cards and modems. The communication sections perform communication processing via networks such as the Internet. Drives are also connected to the I / O interface as needed. Removable media, such as disks, optical discs, magneto-optical discs, semiconductor memories, etc., are installed on the drive as needed so that computer programs read from them can be installed into the storage section as required.

[0109] Specifically, according to embodiments of this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program including a computer program for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication component, and / or installed from a removable medium. When the computer program is executed by a central processing unit (CPU), it performs various functions defined in the system of this application.

[0110] It should be noted that the computer-readable medium shown in the embodiments of this application can be a computer-readable signal medium or a computer-readable storage medium, or any combination of the two. A computer-readable storage medium can be, for example, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), flash memory, optical fiber, portable compact disc read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this application, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying a computer-readable computer program. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media can also be any computer-readable medium other than computer-readable storage media, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The computer program contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to wireless, wired, etc., or any suitable combination thereof.

[0111] The units described in the embodiments of this application can be implemented in software or hardware, and the described units can also be located in a processor. The names of these units do not necessarily limit the specific unit itself.

[0112] Another aspect of this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the voice recognition method for an elevator entrapment scenario as described above. This computer-readable storage medium may be included in the electronic device described in the above embodiments, or it may exist independently and not assembled into the electronic device.

[0113] Another aspect of this application provides a computer program product or computer program including computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform a voice recognition method for an elevator entrapment scenario provided in the various embodiments above.

[0114] The above embodiments are merely illustrative of the principles and effects of the present invention and are not intended to limit the invention. Any person skilled in the art can modify or alter the above embodiments without departing from the spirit and scope of the present invention. Therefore, all equivalent modifications or alterations made by those skilled in the art without departing from the spirit and technical concept disclosed in the present invention should still be covered by the claims of the present invention.

Claims

1. A speech recognition method for elevator entrapment scenarios, characterized in that, include: Acquire voice data in an elevator scenario and preprocess the voice data to obtain a first voice feature; The first speech feature is used as the input to a deep neural network, and the output is the recognized text and the probability value of trapping a person. Based on the identified text and the entrapment probability value, an entrapment determination is made, and an entrapment determination result is output. Specifically, when the entrapment probability value is less than or equal to a first threshold, the entrapment determination result is that no entrapment event has occurred. When the probability value of entrapment is greater than the first threshold and less than the second threshold, keyword matching is performed on the identified text. When a keyword exists, the entrapment determination result is that an entrapment event has occurred; when no keyword exists, the entrapment determination result is that no entrapment event has occurred. When the probability value of entrapment is greater than the second threshold, the entrapment determination result is that an entrapment event has occurred.

2. The speech recognition method for elevator entrapment scenarios according to claim 1, characterized in that, The preprocessing includes: The duration of the voice data is obtained, and the voice data is segmented according to a preset time interval to obtain segmentation features; The frequency of the speech data is obtained, and the feature length of the segmented feature is calculated based on the time interval and the frequency. The first speech feature is obtained by combining the feature length of the segmented feature and the number of segments.

3. The speech recognition method for elevator entrapment scenarios according to claim 1, characterized in that, The deep neural network is a convolutional neural network.

4. The speech recognition method for elevator entrapment scenarios according to claim 3, characterized in that, The step of using the first speech feature as input to a deep neural network and outputting the recognized text and the probability value of trapping the person specifically includes: The first speech feature is used as input and fed into the first convolutional layer for calculation to obtain the second speech feature. The second speech feature is used as input and fed into a four-layer transformer for encoding to obtain the output code. The output encoding is decoded according to a preset dictionary to obtain the recognized text; The output code is used as input to the second convolutional layer for calculation to obtain the output code features; The output encoded features are used as input and fed into two fully connected layers for calculation to obtain feature values ​​with a dimension of one. The feature value is input into a preset trapping probability function to obtain the trapping probability value.

5. The speech recognition method for elevator entrapment scenarios according to claim 4, characterized in that, Decoding the output encoding according to a preset dictionary to obtain the recognized text specifically includes: A probability matrix is ​​generated based on the dictionary, wherein each row vector of the probability matrix contains 3001 probability values, the first 3000 probability values ​​correspond to each Chinese character in the dictionary, and the 3001st probability value corresponds to the empty character; Based on the decoding and the probability matrix, the position with the highest probability value is obtained, that is, the character with the highest probability is found, and the output is the recognized character; Repeat until all the recognized characters in the encoding are obtained, and perform post-processing on the recognized characters to generate recognized text. The post-processing includes removing duplicate characters among adjacent characters and removing empty characters.

6. The speech recognition method for elevator entrapment scenarios according to claim 4, characterized in that, The probability function for trapping people is: ； in, Let be the probability function for trapping people. These are the eigenvalues.

7. A voice recognition system for elevator entrapment scenarios, used to implement the voice recognition method for elevator entrapment scenarios as described in any one of claims 1-6, characterized in that, include: Data acquisition module: Acquires voice data in the elevator scenario and preprocesses the voice data to obtain the first voice feature; Output information module: The first speech feature is used as the input of the deep neural network, and the output is the recognized text and the probability value of trapping people; Comprehensive Judgment Module: Based on the identified text and the entrapment probability value, it determines whether someone is entrapped and outputs the entrapment judgment result; Specifically, this includes: when the probability value of entrapment is less than or equal to the first threshold, the entrapment determination result is that no entrapment event has occurred; When the probability value of entrapment is greater than the first threshold and less than the second threshold, keyword matching is performed on the identified text. When a keyword exists, the entrapment determination result is that an entrapment event has occurred; when no keyword exists, the entrapment determination result is that no entrapment event has occurred. When the probability value of entrapment is greater than the second threshold, the entrapment determination result is that an entrapment event has occurred.

8. An electronic device, characterized in that, The electronic device includes: One or more processors; A storage device for storing one or more programs, which, when executed by one or more processors, cause the electronic device to implement a voice recognition method for an elevator entrapment scenario as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, It stores a computer program, which, when executed by the computer's processor, causes the computer to perform a voice recognition method for an elevator entrapment scenario as described in any one of claims 1 to 6.

Images