Urban Rail Transit Finger and Mouth Call System Based on Posture Recognition

Abstract

The present invention discloses an urban rail transit finger-and-voice call system based on gesture recognition, which includes an acquisition and analysis module for obtaining type A instructions representing indication operations and voice call content, and type B instructions representing actuator operations and voice call content. For the real-time type A instructions of the user, acquisition points representing the time when the user's finger appears in the target area are obtained. Starting from the acquisition point, within the standard duration, the time point when the recognized user's indication operation is correct is marked as the capture moment. The standard duration refers to the duration required for demonstrating the indication operation; it relates to the technical field of urban rail transit. By separately monitoring the driver's indication operations and actuator operations in different ways, irrelevant hand movements in the finger-and-voice call command can be effectively filtered out, avoiding the influence of irrelevant hand movements on the judgment of the finger-and-voice call command, and improving the accuracy of action recognition.

Description

Technical Field

[0001] The present invention belongs to the technical field of urban rail transit, and specifically relates to an urban rail transit finger-and-voice call system based on gesture recognition. Background Art

[0002] Finger-and-voice call in urban rail transit is a working method in the operation of urban rail transit, in which staff point to relevant equipment or signs with their fingers and simultaneously shout relevant content out loud to ensure the accuracy and safety of operations.

[0003] In the prior art, usually during the entire train operation process, it is necessary to monitor the finger-and-voice call commands made by the driver in real time. Continuous operation will not only cause unnecessary energy consumption, but may also accelerate the aging of equipment due to long-term operation. And during the monitoring process, it is very easy to misjudge the finger movements in the finger-and-voice call commands, affecting the judgment result. At the same time, there will also be a problem that the finger movement and the voice call content are correct but there is a delay between the two, without considering the coherence between the finger movement and the voice call content. Furthermore, there is an incorrect judgment on the standardization of the finger-and-voice call commands, resulting in a problem of poor recognition accuracy.

[0004] To solve the above problems, the present invention proposes a solution. Summary of the Invention

[0005] The present invention aims to solve the problems raised in the above background art; for this purpose, the present invention proposes an urban rail transit finger-and-voice call system based on gesture recognition, including:

[0006] An acquisition and analysis module, configured to obtain type A instructions representing the pointing operation and the voice call content, and type B instructions representing the actuator operation and the voice call content;

[0007] And for the real-time type A instructions of the user, obtain the acquisition point representing the time when the user's finger appears in the target area. Starting from the acquisition point, within the standard duration, mark the time point when the recognized user's pointing operation is correct as the capture moment. The standard duration refers to the duration required for demonstrating the pointing operation;

[0008] For type B instructions, when it is detected that the digital mark of the execution signal generated by the sensor is consistent with the digital mark corresponding to the actuator operation in the repository, obtain the generation time of the execution signal;

[0009] Mark the duration between the capture moment and the output time as the interval duration SC1; mark the duration between the generation time of the execution signal and the output time as the interval duration SC2. The output time refers to the time point when it is determined that the driver's voice call content is consistent with the standard text. By comparing the interval duration SC1 and the interval duration SC2 with the preset value AJ respectively, it is used to determine whether the type A instructions and the type B instructions meet the standard specifications.

[0010] Preferably, it further includes an acquisition module for acquiring the number of all stations on the fixed track line from the subway official website.

[0011] Preferably, it further includes a division module for dividing the monitoring section and the acquisition interval according to the number of stations. The specific division method is as follows:

[0012] During the operation of the fixed track line, the total time consumed for the subway to depart from the first station until it leaves the next station is marked as the monitoring section. In the same way as above, according to the number of stations, the number of monitoring sections on the fixed track line is obtained;

[0013] Select any monitoring section, obtain all the finger-and-voice commands within this monitoring section, select any finger-and-voice command, obtain the standard time point when this finger-and-voice command is made, and set the acquisition moment and the end moment before and after the standard time point respectively. Among them, the time intervals between the acquisition moment and the end moment and the standard time point are the same. The time period from the acquisition moment to the end moment is uniformly marked as the acquisition interval;

[0014] In the same way as above, the acquisition intervals of each finger-and-voice command are obtained one by one.

[0015] Preferably, it further includes a preprocessing module for dividing the finger-and-voice commands into type A instructions and type B instructions.

[0016] Preferably, for the indicating operations in type A instructions, the hand area of the driver is photographed by a high-definition camera to obtain the irregular image after shooting, and then the center points of the wrist joint and the elbow joint in the irregular image are determined through a detection algorithm;

[0017] Select any indicating operation, obtain the indicating target in the cab under this indicating operation, and divide the target area according to the indicating target. A three-dimensional rectangular coordinate system is established in the cab with a preset point as the origin. By obtaining the coordinates of all pixel points in the target area, the average value of all coordinates is directly calculated, and then the center point of the target area is determined;

[0018] In the same way as above, the center points of the target areas under each indicating operation are obtained one by one, and an infrared sensor is installed at the position of each center point, and each infrared sensor is marked with a different letter;

[0019] The driver pre-performs the corresponding demonstration indicating operations according to each indicating operation. According to the synchronization principle of the indicating operation and the voice call content, the standard duration of each demonstration indicating operation is obtained through a timer, and the standard durations are stored in the first database in a one-to-one correspondence with the indicating operations.

[0020] Preferably, for the actuator operations in Class B instructions, sensors are pre-set on each actuator, and the execution signals output by each sensor are marked with different numbers and stored in a repository;

[0021] For each actuator operation or instruction operation, the driver uses a voice recognition device to respectively input the oral call content corresponding to the actuator operation or instruction operation, and then converts the oral call content into text information, re-calibrates it into a standard text, and stores the standard text in the second database in a one-to-one correspondence with the actuator operation or instruction operation.

[0022] Preferably, the specific analysis method in the acquisition and analysis module is as follows:

[0023] Optionally select the finger oral call command in a collection interval. For Class A instructions, when the driver's finger points to the target area, the infrared sensor with the corresponding letter mark in the target area will detect a signal change. Then it is judged that the driver's finger movement appears in the target area, and the time point of the first detected signal change is marked as the collection point;

[0024] Furthermore, obtain the center points of the driver's wrist joint and elbow joint, and connect the two to obtain a reference line. Taking the center point of the wrist joint as an end point, extend the reference line infinitely in the direction of the finger. On the edge contour line of the target area, obtain the position of the point with the farthest straight-line distance from the indication point, mark it as the fixed point, and obtain the straight-line distance ZX1 between the fixed point and the indication point;

[0025] Obtain the standard duration corresponding to the indication operation. Starting from the collection point, within the standard duration, obtain the vertical distance CZ from the indication point in the target area to the reference line. When n consecutive cycles all satisfy CZ ≤ ZX1, the judgment result is output, that is, the indication operation is correct, and the moment when the judgment result is output is marked as the capture moment.

[0026] Preferably, for Class B instructions, when the driver performs an actuator operation, the sensor on the corresponding actuator will generate an execution signal. Obtain the digital mark corresponding to the execution signal and the generation time of the execution signal, re-calibrate the obtained digital mark as a to-be-recognized mark, extract the digital mark under the corresponding actuator operation in the repository, re-calibrate it as a specified mark, and compare the to-be-recognized mark with the specified mark. When the to-be-recognized mark does not completely match the specified mark or no execution signal is detected, it is determined that the actuator operation is incorrect. When the to-be-recognized mark completely matches the specified mark, it is determined that the actuator operation is correct.

[0027] Preferably, mark the interval durations between the capture moment, the execution signal generation time and the output time as SC1 and SC2 respectively;

[0028] When SC1 ≤ AJ, it is determined that the Class A instructions completed by the driver meet the specification standards. When SC2 ≤ AJ, it is determined that the Class B instructions completed by the driver meet the specification standards. Conversely, when SC1 > AJ, it is determined that the Class A instructions completed by the driver do not meet the specification standards. When SC2 > AJ, it is determined that the Class B instructions completed by the driver do not meet the specification standards;

[0029] Perform the same processing on the finger-mouth call commands in the remaining acquisition intervals in the above manner, and mark the finger-mouth call commands that do not meet the specification standards as non-standard commands.

[0030] Preferably, it further includes a post-processing module, and the post-processing module is used to process non-standard commands. The specific processing method is as follows:

[0031] On the fixed-rail operation line, obtain the total number A of finger-mouth call commands. For each finger-mouth call command, obtain the total number of times ZCSi that n drivers are marked as non-standard commands under this finger-mouth call command, where i = 1,..., A;

[0032] When ZCSi ≥ Y, it is determined that the probability of the finger-mouth call command not meeting the specification standards is relatively high, where Y is a preset value.

[0033] Compared with the prior art, the beneficial effects of the present invention are:

[0034] The finger-mouth call commands are divided into Class A instructions and Class B instructions. The Class A instructions are the indicating operations and mouth call contents, and the Class B instructions are the actuator operations and mouth call contents. By separately monitoring the indicating operations and actuator operations of the driver in different ways, it is possible to effectively screen out the irrelevant hand movements in the finger-mouth call commands, avoid the irrelevant hand movements from affecting the judgment of the finger-mouth call commands, and improve the accuracy of action recognition;

[0035] At the same time, in the same finger-mouth call command, when it is recognized that the mouth call content of the driver is correct and corresponds one by one to the indicating operation or actuator operation, by analyzing the interval duration between the mouth call content and the indicating operation or actuator operation, the normativity of this finger-mouth call command is judged. And on the fixed-rail line for multiple drivers, by counting the total number of times of being marked as non-standard commands under the same finger-mouth call command, it is possible to obtain the finger-mouth call commands with a relatively high probability of non-standard commands, which is convenient for subsequent targeted training. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 It is a schematic diagram of the principle framework of the present invention;

[0037] Figure 2 It is a schematic diagram of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0038] The technical solution of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0039] Embodiment 1

[0040] Please refer to Figure 1 、 2 , this application provides an urban rail transit finger-and-mouth call system based on gesture recognition, including:

[0041] An acquisition module, which acquires the official website of the subway in the city. The official website of the subway records in detail the line map, station list, transfer guide, and real-time train arrival time information. According to the above information, the number of all stations on the fixed-section track line is obtained, and the number of stations is transmitted to the division module;

[0042] A division module, which receives the number of stations and divides the monitoring section and the acquisition interval according to the number of stations. The specific division method is as follows:

[0043] During the operation of the fixed track line, the total time consumed for the subway to start from the first station until it leaves the next station is marked as the monitoring section. In the same way as above, according to the number of stations, the number of monitoring sections on the fixed track line is obtained;

[0044] Select any monitoring section, and obtain all the finger-and-mouth call commands in this monitoring section. The finger-and-mouth call command refers to the finger action and its corresponding oral content. Select any finger-and-mouth call command, obtain the standard time point for making this finger-and-mouth call command, and set the acquisition moment and the end moment before and after the standard time point respectively. Among them, the time intervals between the acquisition moment and the end moment and the standard time point are the same. The time period from the acquisition moment to the end moment is uniformly marked as the acquisition interval. For each acquisition interval, when the time reaches a preset value near the acquisition moment, the buzzer is started to remind the driver to get ready. When the acquisition moment arrives, the monitoring device is immediately started to start the comprehensive monitoring work. Once the end moment is reached, the monitoring device is turned off and the monitoring task is stopped. There is no need for the monitoring device to be turned on in real time, and it only needs to be enabled during specific periods, which can not only accurately obtain key information but also reduce the operation burden;

[0045] In the same way as above, the acquisition intervals of each finger-and-mouth call command are obtained one by one;

[0046] Preprocessing module, which installs monitoring devices at appropriate positions in the subway cab in advance. The monitoring devices include a high-definition camera and a voice recognition device. The high-definition camera ensures clear shooting of the driver's hand area, and the voice recognition device selects high-performance products to ensure accurate identification of the driver's oral commands and quickly convert them into accurate text information. Before going on duty, the driver establishes a standardized standard for different types of finger-and-voice commands according to the standardization principle of finger-and-voice commands. The specific processing method is as follows:

[0047] The finger-and-voice commands are pre-divided into Class A commands and Class B commands. Class A commands are indication operations and oral commands. For example, when the subway passes through a turnout, point to the turnout position and say: "The turnout is set to straight / lateral, correct." Class B commands are actuator operations and oral commands. For example, when the subway arrives at a station, operate the door control button and say: "The door close button is pressed, and the door is closing." Here, the actuator refers to the instrument and equipment used in Class B commands, including control buttons, operating handles, etc.;

[0048] For the indication operations in Class A commands, the high-definition camera takes pictures of the driver's hand area to obtain an irregular image after shooting. Then, the center points of the wrist joint and elbow joint in the irregular image are determined through a detection algorithm. The method of determining the center point of the irregular image is a prior art, including edge detection method, centroid method, and segmentation method, which will not be described in detail in this application document;

[0049] Taking one of the indication operations as an example, obtain the indication target in the cab under this indication operation, and divide the target area according to the indication target. The target area is preset manually and it is necessary to ensure that the indication target is within the target area. Establish a three-dimensional rectangular coordinate system with a preset point as the origin in the cab. By obtaining the coordinates of all pixel points in the target area, the average value of all coordinates is directly calculated, and then the center point of the target area is determined;

[0050] In the same way as above, obtain the center points of the target areas under each indication operation one by one, re-mark them as indication points, and install infrared sensors at the positions of the indication points. Each infrared sensor is marked with a different letter. The infrared sensor can be of model JHW2-TOC. Of course, those skilled in the art can select other models according to requirements;

[0051] The driver makes corresponding demonstration indication operations according to each indication operation in advance. According to the synchronization principle of indication operation and oral command, the standard duration of each demonstration indication operation is obtained through a timer, and the standard duration is stored in the first database in a one-to-one correspondence with the indication operation;

[0052] For the actuator operations in Class B instructions, sensors are pre-set on each actuator, and the execution signals output by each sensor are marked with different numbers and stored in a repository;

[0053] For each actuator operation or indicated operation, the driver uses a voice recognition device to input the corresponding oral commands for the actuator operation or indicated operation respectively, and then converts the oral commands into text information, re-calibrates it as a standard text, and stores the standard text in a second database in a one-to-one correspondence with the actuator operation or indicated operation;

[0054] A collection and analysis module, during the operation of the subway, for Class A instructions and Class B instructions, collects and analyzes the driver's finger movements and oral commands in real time. The specific analysis methods are as follows:

[0055] Taking the finger and oral commands in one collection interval as an example, for Class A instructions, when the driver's finger points to the target area, the infrared sensor with the corresponding letter mark in the target area will detect a signal change. The signal change is referred to as a signal interruption or weakening, then it is determined that the driver's finger movement appears in the target area, and the time point of the first detected signal change is marked as the collection point;

[0056] Furthermore, obtain the center points of the driver's wrist joint and elbow joint, and connect the two to obtain a reference line. Taking the center point of the wrist joint as an endpoint, extend the reference line infinitely in the direction of the finger. On the edge contour line of the target area, obtain the position of the point with the farthest straight-line distance from the indicated point, mark it as the fixed point, and obtain the straight-line distance ZX1 between the fixed point and the indicated point;

[0057] Obtain the standard duration of the corresponding indicated operation. Starting from the collection point, within the standard duration, obtain the vertical distance CZ from the indicated point in the target area to the reference line. When n consecutive cycles all satisfy CZ ≤ ZX1, then output the judgment result, that is, the indicated operation is correct, and mark the moment of outputting the judgment result as the capture moment;

[0058] Then, use the voice recognition device to recognize the driver's oral commands and convert them into text information. Obtain the standard text corresponding to the indicated operation in the second database, and compare the text information with the standard text;

[0059] When the text information is exactly the same as the standard text, obtain the output time of the comparison result, and obtain the interval duration SC1 between the capture moment and the output time;

[0060] When SC1 ≤ AJ, it is determined that the Class A instruction completed by the driver meets the standard specifications. On the contrary, when SC1 > AJ, it is determined that the Class A instruction completed by the driver does not meet the standard specifications;

[0061] For Class B instructions, when the driver operates the actuator, the sensor on the corresponding actuator generates an execution signal. Obtain the corresponding digital mark on the execution signal and the generation time of the execution signal. Re-calibrate the obtained digital mark as the mark to be recognized. Extract the digital marks corresponding to the actuator operations in the repository and re-calibrate them as specified marks. Compare the mark to be recognized with the specified mark. When the mark to be recognized does not exactly match the specified mark, it is determined that the actuator operation is incorrect. When no execution signal is detected, it is also determined that the actuator operation is incorrect. When the mark to be recognized exactly matches the specified mark, it is determined that the actuator operation is correct;

[0062] Then, use the voice recognition device to recognize the driver's oral command and convert it into text information. Obtain the standard text corresponding to the actuator operation in the second database and compare the text information with the standard text;

[0063] When the text information exactly matches the standard text, obtain the output time of the comparison result and the interval duration SC2 between the generation time of the execution signal and the output time;

[0064] When SC2 ≤ AJ, it is determined that the Class B instruction completed by the driver meets the standard. Conversely, when SC2 > AJ, it is determined that the Class B instruction completed by the driver does not meet the standard, where AJ is the preset duration threshold;

[0065] Process the finger and oral commands in the remaining acquisition intervals in the same way as above, and mark the finger and oral commands that do not meet the standard as non-standard commands.

[0066] Embodiment 2

[0067] In the specific implementation process of this embodiment, compared with Embodiment 1, the specific difference is:

[0068] The post-processing module receives non-standard commands, obtains the finger and oral commands with a relatively high probability of non-standard commands and processes them. The specific processing method is as follows:

[0069] On the fixed-rail operation line, obtain the total number A of finger and oral commands. For each finger and oral command, obtain the total number of times ZCSi that n drivers are marked as non-standard commands under this finger and oral command, i = 1,..., A;

[0070] When ZCSi ≥ Y, it is determined that the probability of non-standard commands for the finger and oral command is relatively high, and the finger and oral commands that meet the above conditions need to be trained intensively later, where Y is a preset value.

[0071] Some of the data in the above formula are numerically calculated after removing their dimensions, and the content not described in detail in this specification belongs to the prior art well-known to those skilled in the art.

[0072] The above embodiments are only used to illustrate the technical method of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical method of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical method of the present invention.

Claims

1. An urban rail transit finger-and-mouth call system based on gesture recognition, characterized in that, Including: A collection and analysis module, configured to obtain type-A instructions representing an indication operation and a voice call content, and type-B instructions representing an actuator operation and a voice call content; And for the real-time type-A instructions of the user, obtain a collection point representing the time when the user's finger appears in the target area. Starting from the collection point, within a standard duration, mark the time point when the recognized user indication operation is correct as the capture moment. The standard duration refers to the duration required for demonstrating the indication operation; For type-B instructions, when it is detected that the digital mark of the execution signal generated by the sensor is consistent with the digital mark under the corresponding actuator operation in the repository, obtain the generation time of the execution signal; Mark the duration between the capture moment and the output time as the interval duration SC1; mark the duration between the generation time of the execution signal and the output time as the interval duration SC2. The output time refers to the time point when it is determined that the driver's voice call content is consistent with the specification text. By comparing the interval duration SC1 and the interval duration SC2 with the preset value AJ respectively, it is used to determine whether the type-A instructions and the type-B instructions meet the specification standards; Optionally select an indication operation, obtain the indication target in the cab under this indication operation, and divide the target area according to the indication target. Establish a three-dimensional rectangular coordinate system with a preset point as the origin in the cab. By obtaining the coordinates of all pixel points in the target area, directly calculate the average value of all coordinates, and then determine the center point of the target area; In the same way as above, obtain the center point of the target area under each indication operation one by one, re-mark it as the indication point, and install an infrared sensor at the position of this center point, and each infrared sensor is marked with a different letter; The specific analysis method in the collection and analysis module is as follows: Optionally select a finger-voice call command in a collection interval. For type-A instructions, when the driver's finger points to the target area, at this time, the infrared sensor corresponding to the letter mark of the target area will detect a signal change, then it is judged that the driver's finger movement appears within this target area, and mark the time point of the first detected signal change as the collection point; Furthermore, obtain the center point of the driver's wrist joint and elbow joint, and connect the two to obtain a reference line. Taking the center point of the wrist joint as an end point, extend the reference line infinitely in the direction of the finger. On the edge contour line of this target area, obtain the position of the point with the farthest straight-line distance from the indication point, mark it as the fixed point, and obtain the straight-line distance ZX1 between the fixed point and the indication point; Obtain the standard duration corresponding to the indication operation. Starting from the collection point, within the standard duration, obtain the vertical distance CZ from the indication point in the target area to the reference line. When it satisfies CZ ≤ ZX1 for consecutive n cycles, output the judgment result, that is, this indication operation is correct, and mark the moment when the judgment result is output as the capture moment.

2. The urban rail transit finger-and-mouth call system based on gesture recognition according to claim 1, wherein, It further includes an acquisition module, configured to obtain the number of all stations on the fixed rail line from the subway official website.

3. The urban rail transit finger and mouth call system based on gesture recognition according to claim 2, characterized in that, It further includes a division module, configured to divide the monitoring section and the collection interval according to the number of stations. The specific division method is: During the operation on a fixed rail line, the total time taken for the subway to start from the first station and leave the next station is marked as the monitoring section. In the same way as above, the number of monitoring sections on the fixed rail line is obtained according to the number of stations. Arbitrarily select a monitoring section, obtain all the finger-and-voice commands within this monitoring section, arbitrarily select a finger-and-voice command, obtain the standard time point for making this finger-and-voice command, and set the acquisition moment and the end moment before and after the standard time point respectively. Among them, the time intervals between the acquisition moment and the end moment and the standard time point are the same. The time period covered from the acquisition moment to the end moment is uniformly marked as the acquisition interval. In the same way as above, obtain the acquisition interval for each finger-and-voice command one by one.

4. The urban rail transit finger and mouth call system based on gesture recognition according to claim 1, characterized in that, It also includes a preprocessing module for classifying finger-and-voice commands into type A instructions and type B instructions.

5. The finger-and-mouth call system for urban rail transit based on gesture recognition according to claim 4, wherein For the indicating operations in type A instructions, the hand area of the driver is photographed by a high-definition camera to obtain the irregular image after shooting, and then the center points of the wrist joint and elbow joint in the irregular image are determined through a detection algorithm. The driver pre-performs corresponding demonstration indicating operations according to each indicating operation. According to the synchronization principle of the indicating operation and the voice call content, the standard duration of each demonstration indicating operation is obtained through a timer and stored in the first database in a one-to-one correspondence with the indicating operation.

6. The urban rail transit finger-mouth calling system based on gesture recognition according to claim 1, wherein, For the actuator operations in type B instructions, sensors are pre-set on each actuator, and the execution signals output by each sensor are marked with different numbers and stored in the storage library. For each actuator operation or indicating operation, the driver uses a voice recognition device to respectively input the voice call content corresponding to the actuator operation or indicating operation, and then converts the voice call content into text information, re-calibrates it as a standard text, and stores the standard text in the second database in a one-to-one correspondence with the actuator operation or indicating operation.

7. The finger-and-mouth calling system for urban rail transit based on gesture recognition according to claim 1, wherein, For type B instructions, when the driver performs an actuator operation, the sensor on the corresponding actuator will generate an execution signal. Obtain the digital mark corresponding to the execution signal and the generation time of the execution signal. Re-calibrate the obtained digital mark as a to-be-recognized mark, extract the digital mark corresponding to the actuator operation in the storage library, and re-calibrate it as a specified mark. Compare the to-be-recognized mark with the specified mark. When the to-be-recognized mark does not completely match the specified mark or no execution signal is detected, it is determined that the actuator operation is incorrect. When the to-be-recognized mark completely matches the specified mark, it is determined that the actuator operation is correct.

8. The urban rail transit finger-and-mouth call system based on gesture recognition according to claim 7, wherein, Mark the time intervals between the capture moment, the generation time of the execution signal and the output time as SC1 and SC2 respectively. When SC1≤AJ, it is determined that the type A instructions completed by the driver meet the standard specifications. When SC2≤AJ, it is determined that the type B instructions completed by the driver meet the standard specifications. On the contrary, when SC1>AJ, it is determined that the type A instructions completed by the driver do not meet the standard specifications. When SC2>AJ, it is determined that the type B instructions completed by the driver do not meet the standard specifications. Perform the same processing on the finger oral commands in the remaining acquisition intervals in the above manner, and mark the finger oral commands that do not meet the specification standards as non-standard commands.

9. The finger-and-mouth call system for urban rail transit based on gesture recognition according to claim 8, wherein It further includes a post-processing module for processing non-standard commands. The specific processing method is as follows: On the fixed-rail operation line, obtain the total number A of finger oral commands. For each finger oral command, obtain the total number of times ZCSi that n drivers are marked as non-standard commands under this finger oral command, where i = 1,..., A; When ZCSi ≥ Y, it is determined that the probability of the finger oral command not meeting the specification standards is relatively high, where Y is a preset value.

Images