A system and method for detecting a single interest behavior of an autistic child

Abstract

The application discloses an autism child single interest behavior detection system, which comprises a test area, test props for setting the area, a visual acquisition device, a voice acquisition device and a terminal; the visual range of the visual acquisition device covers at least the area where the test props are located, and the visual acquisition device is used for acquiring image information; the voice acquisition device is used for acquiring voice information; the terminal communicates with the visual acquisition device and the voice acquisition device; the terminal at least comprises a data storage module for storing data and an algorithm analysis module, which is configured to process the image information by using an object detection algorithm, a hand skeleton point detection algorithm and a visual line estimation algorithm, analyze the state of a subject under different single interest behaviors, the state of the test props and the attention of the subject, determine whether the single interest condition is met, and output a detection result. The application also discloses a detection method using the detection system.

Description

Technical Field

[0001] This invention relates to the field of machine vision, and more particularly to a system and method for detecting single interest behaviors in children with autism. Background Technology

[0002] Autism spectrum disorder (ASD), also known as autism or autistic disorder, is a representative disease of pervasive developmental disorders, mainly characterized by three core symptoms: lack of social interaction skills, lack of communication skills, and restricted interests and repetitive behaviors. Although the exact cause of autism is unknown, there is currently no cure, and no specific drug treatment. However, early detection, early diagnosis, and early intervention—specifically for patients aged 18-36 months—can significantly improve symptoms. Studies have shown that early intervention after early diagnosis can achieve effective recovery in up to 25% of children with ASD.

[0003] Current clinical screening and diagnosis for autism includes a "single interest" behavioral screening method. During the screening process, autistic individuals' "single interest" often manifests as a lack of interest in toys typically enjoyed by other children, but a particular interest in objects not usually considered toys, such as wheels, bottle caps, or other round, rotating objects. Simultaneously, patients often exhibit stereotyped, repetitive actions and unusual behaviors, such as playing with the same toy for extended periods, demanding toys be placed in a fixed location, or consistently playing with toys in the same way. This is one of the more easily observed behavioral characteristics by parents during their child's development and is an important criterion for distinguishing autistic children from typically developing children. The "single interest" behavioral screening method is conducted collaboratively by a qualified physician, parents, and the patient. The physician first communicates with the parents to understand the patient's developmental history. Then, in a designated space, the physician uses testing props to conduct a series of interactions with the patient. Finally, scores are obtained by comparing the results against an autism behavior scale, and the overall assessment yields the diagnostic result. The following disadvantages exist in this manual screening and diagnosis method: (1) The diagnosis process is cumbersome and time-consuming; the manual screening process is complex and requires a long time; (2) The screening process lacks records and is not easy to review; (3) The diagnosis results often have the doctor's subjective opinion and cannot objectively reflect the patient's condition; because doctors are often prone to personal subjective judgment, it affects the patient's condition judgment and the next step of intervention and treatment plan; (4) There are few doctors with professional medical qualifications, and only hospitals in first- and second-tier cities have autism-related diagnostic departments.

[0004] Therefore, those skilled in the art are dedicated to developing a single-interest behavior detection system and method for children with autism, which can simplify the process, reduce time, save manpower and material resources, reduce human interference, and make the detection results more objective. Summary of the Invention

[0005] To achieve the above objectives, the present invention provides a single interest behavior detection system for children with autism, comprising:

[0006] The testing area serves as the testing site;

[0007] The testing equipment is set up within the testing area;

[0008] A visual acquisition device is installed within the test area, and the visual range of the visual acquisition device at least covers the area where the test prop is located; the visual acquisition device is used to acquire image information.

[0009] A voice acquisition device is installed within the test area to collect voice information;

[0010] The terminal communicates with the visual acquisition device and the voice acquisition device to receive the image information and the voice information;

[0011] The terminal includes at least:

[0012] The data storage module is used to store the video information, the voice information, and the detection results;

[0013] The algorithm analysis module is configured to: process the image information using object detection algorithms, hand skeleton point detection algorithms, and gaze estimation algorithms to obtain participant identity information, participant 3D spatial position, facial feature information, human posture information, and voice detection. The facial feature information includes facial feature points and gaze direction, and the human posture information includes human skeleton point information, hand skeleton point information, and head posture information. The module analyzes the subject's state under different single interest behaviors, the state of the test props, and the subject's attention to obtain different concentration time ratios. These concentration time ratios are sorted from largest to smallest. When the difference between the largest and second-largest concentration time ratios exceeds a predetermined threshold, the single interest condition is determined to be met. A quantitative evaluation is performed according to the ratio, and the detection result is output.

[0014] Furthermore, the visual acquisition device includes an RGB camera, a bracket, and a gimbal. The RGB camera is mounted on the gimbal, the gimbal is connected to the bracket, the bracket is set in the test area, and the gimbal is configured to adjust the pitch and yaw angles of the RGB camera.

[0015] Furthermore, the testing equipment includes tables and chairs set up in the testing area, as well as multiple testing toys.

[0016] The present invention also provides a method for detecting a single interest behavior using the single interest behavior detection system described above, the method comprising:

[0017] Step 1: The voice acquisition device detects the tester's voice commands. Once a predetermined command is detected, the terminal begins analysis and testing.

[0018] Step Two: After receiving the terminal prompt, the tester begins the first round of testing for the single interest behavior; the visual acquisition device acquires image information and transmits it to the terminal;

[0019] Step 3: After receiving the terminal prompt, the tester begins the second round of testing for the single interest behavior; the visual acquisition device acquires image information and transmits it to the terminal;

[0020] Step 4: The terminal's algorithm analysis module automatically processes the image information obtained in Step 2 and Step 3, calculates the cumulative score of Step 2 and Step 3, and outputs the result;

[0021] The time interval between steps two and three shall be no less than five minutes.

[0022] Furthermore, step two includes:

[0023] Step 2.1: The tester places all the test toys on the table, and after 2-3 minutes issues the voice command to start the test;

[0024] Step 2.2: Allow the subjects to freely play with the test toys for 8-10 minutes.

[0025] Step 2.3: Using the images acquired by the visual acquisition device, the following visual detection algorithms are used to detect the child's attention: object detection algorithm, hand skeleton point detection algorithm, and gaze estimation algorithm; wherein, the object detection algorithm, hand skeleton point detection algorithm, and gaze estimation algorithm are performed simultaneously.

[0026] Furthermore, step three includes:

[0027] Step 3.1: The tester categorizes the test toys, gives the subject one test toy, and issues the voice command: "Start test," allowing the subject to play freely for 2 minutes.

[0028] Step 3.2: After 2 minutes, the terminal indicates that the test time for this type of toy has ended. The tester then gives the subject the next test toy and issues the voice command: "Start timing, and let the subject play freely for 2 minutes."

[0029] Step 3.3: Using the images acquired by the visual acquisition device, the following visual detection algorithms are used to detect the child's attention: object detection algorithm, hand skeleton point detection algorithm, and gaze estimation algorithm; wherein, the object detection algorithm, hand skeleton point detection algorithm, and gaze estimation algorithm are performed simultaneously;

[0030] Step 3.4: Repeat steps 3.2 and 3.3 until the timer for the last test toy ends.

[0031] Furthermore, the detection process in steps two and three includes three types of events. When the triggering conditions of the three types of events are met, the timing stops and the corresponding video segment is stored.

[0032] Furthermore, the three types of events include:

[0033] Event 1: For the hand skeleton point detection algorithm, the object detection algorithm first obtains the type and coordinates of the test toy held by the subject in real time, obtains the hand skeleton point and coordinates of the subject, calculates the distance between the current test toy and the subject's hand, and starts timing and records a separate video segment when the distance is less than the threshold and exceeds the critical time. When the distance is greater than the maximum threshold and exceeds the critical time, the timing for the test toy is stopped and the video segment is stored separately.

[0034] Event 2: For the gaze estimation algorithm, the gaze vector of the subject is obtained in real time; the gaze origin is the midpoint of the subject's eyes, and the direction is obtained according to the appearance-based gaze estimation algorithm; the object detection algorithm obtains the type and coordinates of the test toy held by the subject in real time, and the center position of the test toy is the center coordinate obtained by object detection; the midpoint of the subject's eyes is connected to obtain the contrast vector; the angle difference between the estimated gaze vector and the contrast vector is calculated; when the angle difference is less than the angle threshold, timing starts, and when the angle difference is greater than the threshold and exceeds the critical time, timing for the test toy stops and the video segment is stored separately.

[0035] Event 3: When Event 1 and Event 2 occur simultaneously, timing and storage of the corresponding video segments begin independently, and storage stops when one of the events ends.

[0036] Furthermore, step four includes:

[0037] Different test toys are numbered, and the video duration corresponding to the trigger events obtained by the visual acquisition device is recorded. The focus time ratio for different test toys is calculated, and the focus time ratios are sorted from largest to smallest. When the difference between the largest focus time ratio and the second largest focus time ratio exceeds a threshold, it is determined that the single interest condition is met, and scores are assigned according to the specified values.

[0038] Furthermore, the threshold is 20%.

[0039] This invention offers the following beneficial technical advantages: It employs an attention detection method that integrates gaze estimation, hand skeleton point detection, and object detection. The main detection process is largely automated by the terminal's algorithm analysis module, with the tester only assisting in the screening process. Data recording and analysis are essentially automated, thus simplifying the process, reducing time, saving manpower and resources, and minimizing human interference. The detection method of this invention provides a standard detection procedure and a method for quantifying the degree of children's "single interest," resulting in more objective detection results.

[0040] The following will further explain the concept, specific structure, and technical effects of the present invention in conjunction with the accompanying drawings, so as to fully understand the purpose, features, and effects of the present invention. Attached Figure Description

[0041] Figure 1 This is a schematic diagram of the detection system structure according to a preferred embodiment of the present invention;

[0042] Figure 2 This is a flowchart of a preferred embodiment of the detection method of the present invention;

[0043] Figure 3 This is a schematic diagram of a hand skeleton point detection algorithm and an object detection algorithm according to a preferred embodiment of the present invention;

[0044] Figure 4 This is a schematic diagram of a preferred embodiment of the line-of-sight estimation algorithm of the present invention;

[0045] Figure 5 This is an event triggering time diagram of a preferred embodiment of the present invention. Detailed Implementation

[0046] The following description, with reference to the accompanying drawings, illustrates several preferred embodiments of the present invention to make its technical content clearer and easier to understand. The present invention can be embodied in many different forms, and the scope of protection of the present invention is not limited to the embodiments mentioned herein.

[0047] In the accompanying drawings, components with the same structure are indicated by the same numerical designation, and components with similar structures or functions are indicated by similar numerical designations. The dimensions and thicknesses of each component shown in the drawings are arbitrary, and the present invention does not limit the dimensions and thicknesses of each component. To make the illustrations clearer, the thickness of some components has been appropriately exaggerated in the drawings.

[0048] like Figure 1 As shown, this invention provides a single-interest behavior detection system for children with autism, including a testing area, testing props, a visual acquisition device, a voice acquisition device, and a terminal. The testing area is the location used for testing, and at least the testing props, visual acquisition device, and voice acquisition device are arranged within the testing area. The terminal can be arranged within the testing area or outside the testing area. The testing area can be a fixed location where the visual acquisition device and testing props are permanently arranged and can be reused. Alternatively, a temporary testing area can be selected, where the testing props, visual acquisition device, voice processing device, and terminal are arranged temporarily after the area is selected.

[0049] In some embodiments, the testing props include at least a test toy used to attract the subject's attention and to observe the subject's behavior and gaze direction while playing with the toy. The test toy can be selected according to actual needs, for example, based on toys known to be of interest to the subject beforehand. The subject playing with the same toy repeatedly can be considered a single interest behavior to be tested. This invention will be described using test toys such as building blocks, ropes, bottle caps, and dolls as examples. Furthermore, for ease of testing, the testing props also include props such as a table 2 and a chair 1 for the subject and tester to sit on, and for placing the test toys.

[0050] In some embodiments, the visual acquisition device includes a high-definition RGB camera 4, a bracket 3, and a pan-tilt head 5. The high-definition RGB camera 4 is mounted on the pan-tilt head 5, which is connected to the bracket 3. The bracket 3 is placed on the floor of the test area or suspended from the ceiling of the test area; this invention uses a tripod as an example. The pan-tilt head 5 can be used to adjust the angle of the high-definition RGB camera 4 so that its field of vision covers the area where the table 2 and chair 1 are placed, allowing for clear and complete image acquisition. The pan-tilt head 5 can be adjusted manually or automatically via a terminal. The pan-tilt head supports at least vertical rotation to adjust the pitch angle and rotation around an axis perpendicular to the ground to adjust the yaw angle. The high-definition RGB camera 4 can be positioned behind and to the right of the test subject's chair; the pitch and yaw angles of the high-definition RGB camera 4 are adjusted via the pan-tilt head 5 to obtain the optimal image; the high-definition RGB camera 4 needs to acquire unobstructed and clear facial information, as well as the patient's hands and the table area.

[0051] In some implementations, the voice acquisition device uses a noise-canceling microphone 7 to obtain high-quality voice from the test subject and control the testing process. The noise-canceling microphone can be miniature and fixed inside the test subject's collar, or it can be placed in the testing area using a support device.

[0052] The terminal communicates with the high-definition RGB camera 4 and the voice acquisition device, receiving information collected by both. The communication method can be any existing technology, such as wireless communication methods like Wi-Fi, Bluetooth, or Zigbee, or wired communication methods like network cables or data cables. The terminal includes at least a data storage module, an algorithm analysis module, and a display device. For example, the terminal can be a PC, with the data storage module and algorithm analysis module integrated in the host computer 8, and the display device being a screen 6 connected to the host computer; alternatively, an integrated terminal can be used, such as an all-in-one machine, a mobile terminal, or an embedded system. The specific form of the terminal does not constitute a limitation of this application.

[0053] The data storage module can be an existing memory in the prior art, used to write video data and voice data from high-definition RGB cameras and voice acquisition devices, facilitating post-processing and diagnostic review by testers.

[0054] The algorithm analysis module can process image information acquired by the visual acquisition device in real time using appearance-based gaze estimation methods and object detection methods. The module utilizes object detection algorithms, gaze estimation algorithms, and hand skeleton point algorithms to process the image information. The processed information mainly includes participant identity information, participant's 3D spatial position, facial feature information, human posture information, and voice detection. Facial feature information includes facial feature points and gaze direction; human posture information includes human skeleton point information, hand skeleton point information, and head posture information. The module analyzes the subject's current state, primarily by assessing the state of the test toy and the subject's attention to determine whether the current state falls under one of the prescribed "single interest" behaviors. The behavior is scored, a total score is calculated, and the final diagnostic result is displayed on screen 6, while this information is also synchronously stored locally. It should be noted that the basis for determining the subject's attention during "single interest" behavior is provided by both gaze and the child-toy relationship.

[0055] like Figure 2As shown, this invention provides a method for detecting "single interest" behavior in autistic individuals, utilizing the aforementioned detection system. Before the test begins, some preparatory work is required. For example, a testing area is prepared, and the subject, accompanied by a supervisor, is guided by the tester to sit in a designated area, for example, on two chairs arranged opposite each other. Before the test, the tester enters the subject's name into the terminal, after which the test can begin. The detection method includes the following steps:

[0056] Step 1: The voice acquisition device detects the tester's voice commands. Once a predetermined command is detected, the terminal begins analysis and testing.

[0057] Step Two: After receiving a prompt from the host, the test subject begins the first round of testing on the "single interest" behavior; the visual acquisition device collects image information and transmits it to the terminal;

[0058] Step 3: After receiving a prompt from the host, the test subject begins the second round of testing on the "single interest" behavior; the visual acquisition device collects image information and transmits it to the terminal;

[0059] Step 4: The terminal's algorithm analysis module automatically processes the image information obtained in Steps 2 and 3, calculates the cumulative score of Steps 2 and 3, and outputs the result;

[0060] The time interval between steps two and three shall be no less than five minutes.

[0061] Step two also includes the following steps:

[0062] Step 2.1: The tester places all the test toys on the table and gives a voice command after 2-3 minutes to start the test;

[0063] Step 2.2: Allow the subjects to freely play with the test toys for 8-10 minutes.

[0064] Step 2.3: Using high-definition images captured by a high-definition RGB camera, the following visual detection algorithms are used to detect children's attention: (1) object detection algorithm; (2) hand skeleton point detection algorithm; (3) gaze estimation algorithm. These three algorithms are performed simultaneously. The detection process can be described as the following three events. Once the trigger condition for each event is met, the timing stops and the corresponding video segment is stored:

[0065] Event 1: such as Figure 3As shown, for the hand skeleton point detection algorithm, the object detection algorithm first obtains the type and coordinates of the test toy held by the subject in real time, obtains the subject's hand skeleton point and coordinates, calculates the distance between the current test toy and the subject's hand, and starts timing and records a separate video segment when the distance is less than the threshold and exceeds the specified "critical time". Then, the timing for the prop is stopped and the video segment is stored separately.

[0066] Event 2: such as Figure 4 As shown, the gaze estimation algorithm is used to obtain the subject's gaze vector in real time; the gaze starting point is the midpoint between the subject's eyes, and the direction is obtained according to the appearance-based gaze estimation algorithm; the object detection algorithm obtains the type and coordinates of the test toy held by the subject in real time, and the center position of the test toy is the center coordinate obtained by the above object detection; the midpoint between the subject's eyes is connected to obtain the contrast vector; the angle difference between the estimated gaze vector and the contrast vector is calculated; when the angle difference is less than the angle threshold, timing starts, and when the angle difference is greater than the threshold and exceeds the specified "critical time", timing for the toy stops and the video segment is stored separately.

[0067] Event 3: such as Figure 5 As shown, when event 1 and event 2 occur simultaneously, timing and storage of the corresponding video segments begin independently, and storage stops when one of the events ends.

[0068] Step three specifically includes the following steps:

[0069] Step 3.1: The tester categorizes the test toys, gives the subject one test toy, and issues a voice command: "Start test," allowing the subject to play freely for 2 minutes.

[0070] Step 3.2: After 2 minutes, the evaluation system will indicate that the time for this type of test toy has ended. The tester will then give the subject the next test toy and issue a voice command: "Start timing." The subject will then play freely for 2 minutes.

[0071] Step 3.3: Same as step 2.3;

[0072] Step 3.4: Repeat steps 3.2 and 3.3 until the timer for the last test toy ends.

[0073] In step four, the terminal's algorithm analysis module processes the image information acquired in steps two and three, and the calculation method is as follows:

[0074]

[0075]

[0076]

[0077] The above formulas are explained as follows:

[0078] The test toys were numbered, with building blocks, rope, bottle caps, and dolls numbered 1, 2, 3, and 4 respectively; the video duration of the trigger event obtained by the high-definition RGB camera was denoted as T. m It is easy to see that the camera acquires video segments x1, x2, x3, and x4 for each type of test toy in the video corresponding to the triggered event. Therefore, for building blocks... The rope has The bottle cap has Dolls

[0079] The above calculation formula can obtain the proportion of attention time R for playing with four toys (playing with one toy can be regarded as a single interest behavior). l (l=1,2,3,4), sort the concentration time proportions from largest to smallest. When the difference between the largest and second largest concentration time proportions exceeds 20%, Δ=R. m -R n If the percentage is >20% (m≠n), it can be determined that the "single interest" condition is met. Based on the difference, the following evaluation scale can be obtained:

[0080] Table 1. Quantification Methods for Children's "Single Interest" Degree

[0081] Focus on the ratio difference Score Δ≤20% 0 20%<Δ≤30% 1 30%<Δ≤40% 2 40%<Δ≤50% 3 Δ>50% 4

[0082] The aforementioned detection method employs an attention detection approach that integrates gaze estimation, hand skeletal point detection, and object detection. The main detection process is largely completed by the terminal's algorithm analysis module, with the tester only playing a supporting role in the screening process. Data recording and analysis are essentially automated, thereby simplifying the process, reducing time, saving manpower and resources, and minimizing human interference. This invention provides a standard detection procedure and a method for quantifying the degree of children's "single interest," resulting in more objective detection results.

[0083] The preferred embodiments of the present invention have been described in detail above. It should be understood that those skilled in the art can make numerous modifications and variations based on the concept of the present invention without creative effort. Therefore, all technical solutions that can be obtained by those skilled in the art based on the concept of the present invention through logical analysis, reasoning, or limited experimentation on the basis of existing technology should be within the scope of protection defined by the claims.

Claims

1. A system for detecting single-interest behaviors in children with autism, characterized in that, include: The testing area serves as the testing site; The testing equipment is set up within the testing area; A visual acquisition device is installed within the test area, and the visual range of the visual acquisition device at least covers the area where the test prop is located. The visual acquisition device is used to acquire image information; A voice acquisition device is installed within the test area to collect voice information; The terminal communicates with the visual acquisition device and the voice acquisition device to receive the image information and the voice information; The terminal includes at least: The data storage module is used to store the image information, the voice information, and the detection results; The algorithm analysis module is configured to process the image information using object detection, hand skeleton point detection, and gaze estimation algorithms to obtain participant identity information, participant's three-dimensional spatial position, facial feature information, human posture information, and voice detection. The facial feature information includes facial feature points and gaze direction, and the human posture information includes human skeleton point information, hand skeleton point information, and head posture information. The object detection, hand skeleton point detection, and gaze estimation algorithms are performed simultaneously. The process of detecting the subject's attention includes three types of events. When each type of event meets its trigger condition, the timing stops, and the corresponding video segment is stored. Event 1: For the hand skeleton point detection algorithm, the object detection algorithm first obtains the type and coordinates of the test toy held by the subject in real time, obtains the hand skeleton point and coordinates of the subject, calculates the distance between the current test toy and the subject's hand, and starts timing and records a separate video segment when the distance is less than the threshold and exceeds the critical time. When the distance is greater than the maximum threshold and exceeds the critical time, the timing for the test toy is stopped and the video segment is stored separately. Event 2: For the gaze estimation algorithm, the gaze vector of the subject is obtained in real time; the gaze origin is the midpoint of the subject's eyes, and the direction is obtained according to the appearance-based gaze estimation algorithm; the object detection algorithm obtains the type and coordinates of the test toy held by the subject in real time, and the center position of the test toy is the center coordinate obtained by object detection; the midpoint of the subject's eyes is connected to obtain the contrast vector; the angle difference between the estimated gaze vector and the contrast vector is calculated; when the angle difference is less than the angle threshold, timing starts, and when the angle difference is greater than the threshold and exceeds the critical time, timing for the test toy stops and the video segment is stored separately. Event 3: When Event 1 and Event 2 occur simultaneously, timing and storage of corresponding video segments begin independently until one of the events ends and storage stops; the state of the subject under different single interest behaviors, the state of the test props, and the subject's attention are analyzed to obtain different concentration time ratios. The concentration time ratios are sorted from largest to smallest. When the difference between the largest concentration time ratio and the second largest concentration time ratio exceeds a predetermined threshold, the single interest condition is determined to be met, and a quantitative evaluation is performed according to the ratio, and the detection result is output.

2. The single interest behavior detection system as described in claim 1, characterized in that, The visual acquisition device includes an RGB camera, a bracket, and a gimbal. The RGB camera is mounted on the gimbal, which is connected to the bracket. The bracket is set in the test area, and the gimbal is configured to adjust the pitch and yaw angles of the RGB camera.

3. The single interest behavior detection system as described in claim 1, characterized in that, The testing equipment includes tables and chairs set up in the testing area, as well as multiple testing toys.

4. The single interest behavior detection system as described in claim 1, characterized in that... The terminal is configured to perform the following operations: Step 1: The voice acquisition device detects the tester's voice commands. Once a predetermined command is detected, the terminal begins analysis and testing. Step Two: After receiving the terminal prompt, the tester begins the first round of testing for the single interest behavior; the visual acquisition device acquires image information and transmits it to the terminal; Step 3: After receiving the terminal prompt, the tester begins the second round of testing for the single interest behavior; the visual acquisition device acquires image information and transmits it to the terminal; Step 4: The terminal's algorithm analysis module automatically processes the image information obtained in Step 2 and Step 3, calculates the cumulative score of Step 2 and Step 3, and outputs the result; The time interval between steps two and three shall be no less than five minutes.

5. The single interest behavior detection system as described in claim 4, characterized in that, Step two includes: Step 2.1: The tester places all the test toys on the table, and after 2-3 minutes issues the voice command to start the test; Step 2.2: Allow the subjects to freely play with the test toys for 8-10 minutes. Step 2.3: Using the images acquired by the visual acquisition device, the following visual detection algorithms are used to detect the child's attention: object detection algorithm, hand skeleton point detection algorithm, and gaze estimation algorithm; wherein, the object detection algorithm, hand skeleton point detection algorithm, and gaze estimation algorithm are performed simultaneously.

6. The single interest behavior detection system as described in claim 4, characterized in that, Step three includes: Step 3.1: The tester categorizes the test toys, gives the subject one test toy, and issues the voice command: "Start test," allowing the subject to play freely for 2 minutes. Step 3.2: After 2 minutes, the terminal indicates that the test toy time is over, and the tester gives the subject the next test toy and issues the voice command: Start timing, and let the subject play freely for 2 minutes; Step 3.3: Using the images acquired by the visual acquisition device, the following visual detection algorithms are used to detect the child's attention: object detection algorithm, hand skeleton point detection algorithm, and gaze estimation algorithm; wherein, the object detection algorithm, hand skeleton point detection algorithm, and gaze estimation algorithm are performed simultaneously; Step 3.4: Repeat steps 3.2 and 3.3 until the timer for the last test toy ends.

7. The single interest behavior detection system as described in claim 4, characterized in that, Step four includes: Different test toys are numbered, and the video duration corresponding to the trigger events obtained by the visual acquisition device is recorded. The focus time ratio for different test toys is calculated, and the focus time ratios are sorted from largest to smallest. When the difference between the largest focus time ratio and the second largest focus time ratio exceeds a threshold, it is determined that the single interest condition is met, and scores are assigned according to the specified values.

8. The single interest behavior detection system as described in claim 7, characterized in that, The threshold is 20%.

Images