A method and system for gait state recognition based on acceleration signals

Abstract

This invention discloses a gait state recognition method and system based on acceleration signals. The method includes: acquiring target axis acceleration signals at the waist position of the human body and performing signal preprocessing to obtain compensated acceleration signals; determining stable human motion state information; performing peak and valley event detection to determine whether the compensated acceleration signals are peak or valley events, and outputting the peak and valley event detection results of the compensated acceleration signals; constructing gait cycles and performing initial classification of gait cycles to obtain original gait types; and performing phased smoothing processing on the original gait types based on stable human motion state information to determine and count the current gait state, thereby realizing gait state recognition of the human body. This invention can reduce hardware complexity and achieve stable recognition and classification of different gait states. As a gait state recognition method and system based on acceleration signals, this invention can be widely applied in the field of human motion recognition technology.

Description

Technical Field

[0001] This invention relates to the field of human motion recognition technology, and in particular to a gait state recognition method and system based on acceleration signals. Background Technology

[0002] With the continuous development of wearable devices, human motion monitoring, and behavior recognition technologies, inertial sensor-based gait analysis has been widely applied in fields such as rehabilitation training, sports health assessment, smart wearables, indoor positioning, and human-computer interaction. Among these, acceleration signals, due to their simple acquisition methods, low hardware cost, and low power consumption, have become an important source of information for gait detection and motion state recognition. By analyzing the temporal changes, periodic characteristics, and amplitude characteristics of acceleration signals during human movement, it is possible to detect human walking behavior, count steps, and determine gait state, thus possessing high application value.

[0003] However, existing gait recognition methods still have certain shortcomings. Some solutions rely heavily on multi-sensor fusion, attitude calculation, or complex installation conditions, resulting in complex system implementation and limited applicability. Others rely solely on a single threshold or simple features for judgment, making them susceptible to noise interference, individual differences, and factors such as starting, stopping, and changes in stride frequency. This leads to unstable switching between static and dynamic states and makes it difficult to accurately distinguish between similar movement patterns such as stepping in place, forward gait, and backward gait. Summary of the Invention

[0004] To address the aforementioned technical problems, the present invention aims to provide a gait state recognition method and system based on acceleration signals, which can reduce hardware complexity and achieve stable recognition and classification of different gait states.

[0005] The first technical solution adopted in this invention is: a gait state recognition method based on acceleration signals, comprising the following steps: The target axis acceleration signal at the waist position of the human body is collected and preprocessed to obtain the compensated acceleration signal. The current dynamic and static state of the human body is detected based on the compensated acceleration signal to determine the stable motion state information of the human body; Peak and valley event detection is performed on the compensated acceleration signal to determine whether the compensated acceleration signal is a peak event or a valley event, and the peak and valley event detection results of the compensated acceleration signal are output. The peak and valley event detection results of the compensated acceleration signal are used to construct the gait cycle, and the gait cycle is initially classified to obtain the original gait type; Based on information about stable human movement, the original gait type is smoothed in stages to obtain the smoothed human gait type. The current gait state is determined and counted based on the smoothed human gait type, thereby realizing the recognition of the human gait state.

[0006] Furthermore, the step of acquiring the target axis acceleration signal at the waist position of the human body and performing signal preprocessing to obtain the compensated acceleration signal specifically includes: Based on preset frequency information, the target axis acceleration signal at the waist position of the human body is continuously sampled by an accelerometer to obtain the original acceleration sequence signal. The original acceleration sequence signal is processed to unify the direction, resulting in a unified original acceleration sequence signal. Based on the preset baseline estimation window length, baseline estimation is performed on the unified original acceleration sequence signal to obtain the baseline value; The difference between the unified original acceleration sequence signal and the baseline value is obtained to obtain the compensated acceleration signal.

[0007] Furthermore, the step of detecting the current dynamic and static state of the human body based on the compensated acceleration signal to determine the stable motion state information of the human body specifically includes: Based on the first preset sliding window, the mean value of the compensated acceleration signal is calculated; Based on the first preset sliding window, and combined with the mean of the compensated acceleration signal, the variance of the compensated acceleration signal is calculated. Based on the window variance distribution of the compensated acceleration signals under preset human static state and walking motion state, the preset motion threshold is determined; The variance of the compensated acceleration signal is compared with a preset motion threshold to determine the current state of the human body, which includes the original motion state and the original stationary state. Based on the second preset sliding window, the current state of the human body is proportionally determined to determine the proportion of human body movement. Combined with the preset movement determination proportion threshold, the stable movement state information of the human body is determined. The stable movement state information of the human body includes the movement initiation stage and the stable movement stage.

[0008] Furthermore, the step of performing peak-valley event detection on the compensated acceleration signal, determining whether the compensated acceleration signal is a peak event or a valley event, and outputting the peak-valley event detection result of the compensated acceleration signal specifically includes: Based on the third preset sliding window, the window mean of the compensated acceleration signal is calculated; Based on the third preset sliding window, and combined with the window mean of the compensated acceleration signal, the average absolute deviation of the compensated acceleration signal is calculated. Based on the window mean of the compensated acceleration signal and the mean absolute deviation of the compensated acceleration signal, an adaptive peak threshold and an adaptive valley threshold are constructed. If the compensated acceleration signal of the current sampling point is greater than the compensated acceleration signals of its adjacent sampling points, and the compensated acceleration signal of the current sampling point is greater than the adaptive peak threshold, then the compensated acceleration signal of the current sampling point is marked as a peak event. If the compensated acceleration signal of the current sampling point is less than the compensated acceleration signals of its adjacent sampling points, and the compensated acceleration signal of the current sampling point is less than the adaptive valley threshold, then the compensated acceleration signal of the current sampling point is marked as a valley event.

[0009] Furthermore, it also includes performing fallback confirmation on the compensated acceleration signal of the current sampling point marked as a peak event, calculating the peak candidate amplitude of the compensated acceleration signal of the current sampling point, and calculating the difference between the compensated acceleration signal and the adjacent previous and next sampling points. When the difference is greater than the fallback confirmation coefficient multiplied by the dynamic noise estimate of the compensated acceleration signal of the current sampling point, it is determined that the fallback confirmation condition is met, and the compensated acceleration signal of the current sampling point is confirmed as a peak event. If the fallback confirmation condition is not met, the peak event is removed.

[0010] Furthermore, the step of constructing gait cycles from the peak and valley event detection results of the compensated acceleration signal, and performing initial classification of the gait cycles to obtain the original gait type, specifically includes: The gait period is constructed by combining the valley event of the compensated acceleration signal at the previous sampling point, the peak event of the compensated acceleration signal at the current sampling point, and the valley event of the compensated acceleration signal at the next sampling point. The rising and falling time-series characteristics of the gait cycle are determined. The rising time-series characteristics are determined by the difference between the peak event of the compensated acceleration signal at the current sampling point and the valley event of the compensated acceleration signal at the previous sampling point. The falling time-series characteristics are determined by the difference between the valley event of the compensated acceleration signal at the next sampling point and the peak event of the compensated acceleration signal at the current sampling point. The characteristic ratio of the gait cycle is determined based on the ratio between the rising and falling time-series characteristics of the gait cycle. The current gait cycle is initially classified based on the characteristic ratio of the gait cycle to obtain the original gait type.

[0011] Furthermore, the step of initially classifying the current gait cycle based on the feature ratio of the gait cycle to obtain the original gait type specifically includes: If the feature ratio of the current gait cycle is greater than the preset forward gait determination threshold, then the original gait type of the current gait cycle is determined to be a forward state. If the feature ratio of the current gait cycle falls within the preset backward gait determination threshold range, then the original gait type of the current gait cycle is determined to be backward gait. If the absolute value of the difference between the feature ratio of the current gait cycle and the feature ratio of the previous valid gait cycle is less than the preset tolerance for stationary gait determination or does not satisfy the forward and backward gait conditions, then the original gait type of the current gait cycle is determined to be stationary gait.

[0012] Furthermore, the step of performing staged smoothing processing on the original gait type based on human stable motion state information to obtain the smoothed human gait type specifically includes: If the stable motion state information of the human body is the initial stage of motion, start voting is performed on multiple consecutive original gait types to obtain the gait type of the initial stage. If the human body's stable motion state information is in the stable motion stage, a steady-state vote is performed on several original gait types to obtain a stable output gait type; By combining the gait type at the initial stage with the gait type at stable output, a smoothed human gait type is constructed.

[0013] Furthermore, the step of determining the current gait state based on the smoothed human gait type and counting it to achieve gait state recognition specifically includes: The smoothed human gait types are voted on to obtain the current gait state; For the stable output gait type in the smoothed human gait type, the total number of steps, forward steps, backward steps and stationary steps are accumulated to obtain the current movement step count statistics; By combining the current gait status with the current number of steps, the gait status of the human body can be identified.

[0014] The second technical solution adopted in this invention is: a gait state recognition system based on acceleration signals, comprising: The first module is used to collect the target axis acceleration signal at the waist position of the human body and perform signal preprocessing to obtain the compensated acceleration signal. The second module is used to detect the current dynamic and static state of the human body based on the compensated acceleration signal and determine the stable motion state information of the human body. The third module is used to perform peak and valley event detection on the compensated acceleration signal, determine whether the compensated acceleration signal is a peak event or a valley event, and output the peak and valley event detection results of the compensated acceleration signal. The fourth module is used to detect peak and valley events in the compensated acceleration signal, construct gait cycles, and perform initial classification of gait cycles to obtain the original gait type. The fifth module is used to perform phased smoothing of the original gait type based on the human body's stable motion state information, so as to obtain the smoothed human gait type. The sixth module is used to determine the current gait state based on the smoothed human gait type and count it, thereby realizing the recognition of the human gait state.

[0015] The beneficial effects of the method and system of this invention are as follows: This invention acquires the target axis acceleration signal at the waist position of the human body and performs signal preprocessing to obtain a compensated acceleration signal; then, based on the compensated acceleration signal, it detects the current dynamic and static state of the human body to determine the stable motion state information of the human body; further, it performs peak and valley event detection on the compensated acceleration signal to determine whether the compensated acceleration signal is a peak event or a valley event, and outputs the peak and valley event detection results of the compensated acceleration signal. This allows for dynamic and static state determination, peak and valley event extraction, and gait state classification based solely on the acceleration signal, reducing the hardware complexity of the system; and further... The peak and valley event detection results of the compensated acceleration signal are used to construct the gait cycle and perform initial classification of the gait cycle to obtain the original gait type. Based on the stable motion state information of the human body, the original gait type is smoothed in stages to obtain the smoothed human gait type. Finally, the current gait state is determined and counted according to the smoothed human gait type to realize the recognition of the human gait state. Through direction unification, baseline compensation, adaptive peak and valley detection, and a voting mechanism that combines the starting stage and steady state stage, the stability during the transition between stationary and moving states is improved, and the ability to distinguish between forward gait, backward gait and stationary gait is enhanced. Attached Figure Description

[0016] Figure 1 This is a flowchart of the steps of a gait state recognition method based on acceleration signals according to the present invention; Figure 2 This is a structural block diagram of a gait state recognition system based on acceleration signals according to the present invention; Figure 3 This is a flowchart illustrating the gait state recognition method based on acceleration signals provided in a specific embodiment of the present invention; Figure 4 This is a schematic diagram of the gait cycle construction and classification process provided in a specific embodiment of the present invention. Detailed Implementation

[0017] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments. The step numbers in the following embodiments are only for ease of explanation and do not limit the order of the steps. The execution order of each step in the embodiments can be adapted according to the understanding of those skilled in the art.

[0018] Reference Figure 1 and Figure 3 This invention provides a gait state recognition method based on acceleration signals, the method comprising the following steps: S100: Collect the target axis acceleration signal at the waist position of the human body and perform signal preprocessing to obtain the compensated acceleration signal; Specifically, based on preset frequency information, the target axis acceleration signal at the waist position of the human body is continuously sampled by an accelerometer to obtain the original acceleration sequence signal; the original acceleration sequence signal is subjected to direction unification processing to obtain the unified original acceleration sequence signal; based on the preset baseline estimation window length, the unified original acceleration sequence signal is baseline estimated to obtain the baseline value; the difference between the unified original acceleration sequence signal and the baseline value is obtained to obtain the compensated acceleration signal.

[0019] In this embodiment, firstly, the accelerometer is worn on the front or back of the waist, and the target axis acceleration signal is continuously sampled at a preset sampling frequency to obtain the original acceleration sequence. The sampling frequency is 100Hz.

[0020] To further address the issue of reversed target axis directions caused by wearing the device in front of the waist versus behind the waist, the original acceleration signal was processed to unify its direction, resulting in a unified acceleration signal: ; in, This is the direction factor; when the sensor is worn in front of the waist, When the sensor is worn on the back of the waist ;or, Determined by the installation calibration results.

[0021] Then, baseline estimation is performed on the unified acceleration signal, and the baseline value is... Calculated using the sliding window mean: ; in, The baseline estimation window length is determined. The sliding window ends at the current sampling point and includes the current sampling point and all preceding points. The system uses 100 sampling points, which are updated point by point as new sampling points are input, to estimate the slowly changing components in the acceleration signal.

[0022] Finally, the baseline value is subtracted from the current acceleration signal to obtain the compensated acceleration signal. : ; The compensated acceleration signal is used for subsequent motion and static detection, peak and valley detection, and gait classification.

[0023] S200: Detect the current dynamic and static state of the human body based on the compensated acceleration signal to determine the stable motion state information of the human body; Specifically, based on a first preset sliding window, the mean of the compensated acceleration signal is calculated; based on the first preset sliding window and the mean of the compensated acceleration signal, the variance of the compensated acceleration signal is calculated; based on the window variance distribution of the compensated acceleration signal under preset human stationary and walking states, a preset motion threshold is determined; the variance of the compensated acceleration signal is compared with the preset motion threshold to determine the current state of the human body, which includes the original motion state and the original stationary state; based on a second preset sliding window, the current state of the human body is proportionally determined to determine the human body motion ratio, and combined with a preset motion determination ratio threshold, the stable motion state information of the human body is determined, which includes the motion initiation stage and the stable motion stage.

[0024] In this embodiment, firstly, the mean value of the compensated acceleration signal is calculated within the first sliding window: ; in, This is the length of the motion / static detection window.

[0025] Furthermore, the variance of the compensated acceleration signal is calculated within the first sliding window: ; Furthermore, the variance With preset motion threshold By comparison, the original motion determination result at the current moment is obtained. : ; in, Indicates the original state of motion. This indicates the original static state. The preset motion threshold... It can be obtained through calibration experiments or preliminary experiments, specifically determined based on the window variance distribution of the compensated acceleration signal in static and walking states, in order to distinguish between small random fluctuations when stationary and periodic amplitude changes when moving.

[0026] To improve the stability of motion-to-static switching, a proportional determination is performed on multiple consecutive original motion determination results within the second sliding window to obtain the motion ratio: ; in, The length of the status voting window. The motion ratio. Indicates that recently The percentage of original motion determination results that are determined to be in motion is used to reflect the overall trend of motion determination results over a continuous period of time near the current moment, rather than determining the motion state based solely on the instantaneous determination result of a single sampling point.

[0027] Finally, a stable motion state is obtained based on the stated motion ratio. : ; in, The threshold for motion determination ratio, Indicates a stable state of motion. It indicates a stable, static state.

[0028] S300: Perform peak and valley event detection on the compensated acceleration signal, determine whether the compensated acceleration signal is a peak event or a valley event, and output the peak and valley event detection result of the compensated acceleration signal. Specifically, based on the third preset sliding window, the window mean of the compensated acceleration signal is calculated; based on the third preset sliding window and combined with the window mean of the compensated acceleration signal, the mean absolute deviation of the compensated acceleration signal is calculated; based on the window mean of the compensated acceleration signal and the mean absolute deviation of the compensated acceleration signal, an adaptive peak threshold and an adaptive valley threshold are constructed; if the compensated acceleration signal of the current sampling point is greater than the compensated acceleration signals of its adjacent sampling points, and the compensated acceleration signal of the current sampling point is greater than the adaptive peak threshold, then the compensated acceleration signal of the current sampling point is marked as a peak event; if the compensated acceleration signal of the current sampling point is less than the compensated acceleration signals of its adjacent sampling points, and the compensated acceleration signal of the current sampling point is less than the adaptive valley threshold, then the compensated acceleration signal of the current sampling point is marked as a valley event.

[0029] In addition, this embodiment also includes performing fallback confirmation on the compensated acceleration signal of the current sampling point marked as a peak event, calculating the peak candidate amplitude of the compensated acceleration signal of the current sampling point, and calculating the difference between the compensated acceleration signal and the adjacent previous and next sampling points. When the difference is greater than the fallback confirmation coefficient multiplied by the dynamic noise estimate of the compensated acceleration signal of the current sampling point, it is determined that the fallback confirmation condition is met, and the compensated acceleration signal of the current sampling point is confirmed as a peak event. If the fallback confirmation condition is not met, the peak event is eliminated.

[0030] In this embodiment, firstly, the window mean of the compensated acceleration signal is calculated within the third sliding window: ; And calculate the mean absolute deviation: ; in, This is the length of the peak-valley detection window.

[0031] Furthermore, adaptive peak thresholds and adaptive valley thresholds are constructed based on the window mean and mean absolute deviation: ; ; in, This is the threshold adjustment coefficient.

[0032] The signal corresponding to the current sampling point is considered a peak candidate event when the following conditions are met: ; ; in, This refers to the sampling sequence number corresponding to the previously confirmed peak event. This represents the minimum peak interval. Among the above conditions, and Indicates sampling point The compensated acceleration signal at a certain point is greater than that at its adjacent sampling points, exhibiting local peak characteristics. This indicates that the local peak amplitude exceeds the adaptive peak threshold, which is used to exclude small noise fluctuations; This is used to limit the minimum time interval between adjacent peak events, and to avoid repeated peak candidate events caused by signal jitter within the same step state period.

[0033] Furthermore, to avoid false detections caused by noise, peak candidate events are confirmed by fallback; when the peak candidate amplitude... It satisfies the following relationship with subsequent signal values: ; The peak event was confirmed, in which, This is the pullback confirmation coefficient.

[0034] The valley event is confirmed when the following conditions are met: ; ; in, This is the sampling sequence number corresponding to the last confirmed valley event. Among the above conditions, and This indicates that the compensated acceleration signal at sampling point n-1 is less than that at its adjacent sampling points before and after it, exhibiting local valley characteristics; This indicates that the amplitude of the local valley value is lower than the adaptive valley value threshold, which is used to exclude small noise fluctuations; Used to limit the minimum time interval between adjacent valley events, avoiding repeated valley events caused by signal jitter within the same step state period.

[0035] The peak and valley event detection results of the S400 and compensated acceleration signals are used to construct the gait cycle and perform initial classification of the gait cycle to obtain the original gait type. Specifically, a gait cycle is constructed by combining the valley event of the compensated acceleration signal at the previous sampling point, the peak event of the compensated acceleration signal at the current sampling point, and the valley event of the compensated acceleration signal at the next sampling point. The rising and falling time-series characteristics of the gait cycle are determined. The rising time-series characteristic is determined by the difference between the peak event of the compensated acceleration signal at the current sampling point and the valley event of the compensated acceleration signal at the previous sampling point. The falling time-series characteristic is determined by the difference between the valley event of the compensated acceleration signal at the next sampling point and the peak event of the compensated acceleration signal at the current sampling point. A feature ratio of the gait cycle is determined based on the ratio between the rising and falling time-series characteristics. The current gait cycle is initially classified based on the feature ratio of the gait cycle to obtain the original gait type.

[0036] Specifically, if the feature ratio of the current gait cycle is greater than the preset forward gait determination threshold, the original gait type of the current gait cycle is determined to be a forward state; if the feature ratio of the current gait cycle falls within the preset backward gait determination threshold range, the original gait type of the current gait cycle is determined to be a backward gait; if the absolute value of the difference between the feature ratio of the current gait cycle and the feature ratio of the previous valid gait cycle is less than the preset stationary gait determination tolerance or does not satisfy the forward and backward gait conditions, the original gait type of the current gait cycle is determined to be a stationary gait.

[0037] In this embodiment, as Figure 4As shown, firstly, according to the "previous valley event" Peak events The next valley event A complete gait cycle is constructed by following the sequence of "".

[0038] Furthermore, the rising and falling time-series characteristics of the gait cycle are calculated: ; ; in, , and These represent the sampling sequence numbers corresponding to the previous valley event, peak event, and subsequent valley event, respectively.

[0039] Then, the characteristic ratio of the gait period is calculated: ; Finally, based on the eigenricular ratio Perform an initial classification of the current gait cycle to obtain the original gait type. : ; in, The characteristic ratio of the previous effective gait cycle. To determine the tolerance for gait in place, The threshold for determining the forward state. and The threshold range for determining backward gait.

[0040] S500: Based on the information of stable human motion state, the original gait type is smoothed in stages to obtain the smoothed human gait type. Specifically, if the stable motion state information of the human body is the initial stage of motion, a starting vote is performed on multiple consecutive original gait types to obtain the starting stage gait type; if the stable motion state information of the human body is the stable motion stage, a steady-state vote is performed on several original gait types to obtain the stable output gait type; combining the starting stage gait type and the stable output gait type, a smoothed human gait type is constructed.

[0041] In this embodiment, firstly, during the initial movement phase, a starting vote is performed on multiple consecutive original gait types to obtain the starting phase gait type. : ; in, To represent the mode operation, This is the starting voting window length.

[0042] Furthermore, during the stable movement phase, a steady-state vote is performed on the most recent primitive gait types to obtain the stable output gait type. : ; in, This represents the steady-state voting window length.

[0043] Finally, when a stable motion state is detected to switch from motion to stillness, if the preset voting length has not been reached in the initial stage, the existing gait cycles are supplemented with statistics using a unified voting method to maintain consistency between the total number of steps and the statistical results of various types of steps.

[0044] S600: Determine the current gait state based on the smoothed human gait type and count it to achieve gait state recognition of the human body.

[0045] Specifically, a state vote is performed on the smoothed human gait types to obtain the current gait state; the total number of steps, forward steps, backward steps, and stationary steps of the stable output gait types in the smoothed human gait types are accumulated to obtain the current movement step count statistics; the current gait state and the current movement step count statistics are combined to realize the recognition of the human gait state.

[0046] In this embodiment, firstly, a state vote is performed on the gait types of the most recent stable outputs to obtain the current gait state. : ; in, This is the length of the status voting window.

[0047] The current gait state includes at least the stationary state, the forward state, the backward state, and the stationary state.

[0048] Furthermore, based on the gait type of stable output, the total number of steps, forward steps, backward steps, and in-place steps are accumulated respectively. Let the indicator function be... Then we have: ; ; ; ; in, Indicates a forward-moving state. Indicates a backward gait. Indicates gait in place. Indicates the total number of steps. Indicates the number of steps to move forward. Indicates the number of steps backward. Indicates the number of steps taken in place. The indicator function... This indicates the result of the conditional judgment; the value is 1 if the condition within the parentheses is true, and 0 otherwise. Therefore, the gait type at which the system obtains stable output each time is determined. Then, the total number of steps is accumulated based on whether it belongs to forward, backward or stationary gait, and the forward step count, backward step count and stationary step count are updated according to its specific type.

[0049] Finally, the current gait state and the corresponding step count are output.

[0050] Finally, it should be noted that the sampling frequency, baseline estimation window length, dynamic and static detection window length, peak and valley detection window length, minimum peak and valley interval, starting voting window length, steady-state voting window length, and state voting window length in the method of this embodiment can all be adjusted according to the wearing position, the user's cadence characteristics, and the application scenario.

[0051] In summary, the embodiments of the present invention can realize dynamic and static state determination, peak and valley event extraction, and gait state classification based solely on acceleration signals, thereby reducing the hardware complexity of the system. At the same time, through direction unification, baseline compensation, adaptive peak and valley detection, and a voting mechanism that combines the starting phase and steady-state phase, the stability during the transition between stationary and moving states is improved, and the ability to distinguish between forward gait, backward gait, and stationary gait is enhanced.

[0052] Reference Figure 2 A gait state recognition system based on acceleration signals, comprising: The first module 201 is used to collect the target axis acceleration signal at the waist position of the human body and perform signal preprocessing to obtain the compensated acceleration signal. The second module 202 is used to detect the current dynamic and static state of the human body based on the compensated acceleration signal and determine the stable motion state information of the human body. The third module 203 is used to perform peak and valley event detection on the compensated acceleration signal, determine whether the compensated acceleration signal is a peak event or a valley event, and output the peak and valley event detection result of the compensated acceleration signal. The fourth module 204 is used to detect peak and valley events of the compensated acceleration signal, construct gait cycles, and perform initial classification of gait cycles to obtain the original gait type; The fifth module 205 is used to perform phased smoothing of the original gait type based on the human stable motion state information to obtain the smoothed human gait type. The sixth module 206 is used to determine the current gait state based on the smoothed human gait type and count it, thereby realizing the recognition of the human gait state.

[0053] The content of the above method embodiments is applicable to this system embodiment. The specific functions implemented in this system embodiment are the same as those in the above method embodiments, and the beneficial effects achieved are also the same as those achieved in the above method embodiments.

[0054] The preferred embodiments of the present application have been described above with reference to the accompanying drawings, but this is not intended to limit the scope of the embodiments of the present application. Any modifications, equivalent substitutions, and improvements made by those skilled in the art without departing from the scope and substance of the embodiments of the present application shall be within the scope of the embodiments of the present application.

Claims

1. A gait state recognition method based on acceleration signals, characterized in that, Includes the following steps: The target axis acceleration signal at the waist position of the human body is collected and preprocessed to obtain the compensated acceleration signal. The current dynamic and static state of the human body is detected based on the compensated acceleration signal to determine the stable motion state information of the human body; Peak and valley event detection is performed on the compensated acceleration signal to determine whether the compensated acceleration signal is a peak event or a valley event, and the peak and valley event detection results of the compensated acceleration signal are output. The peak and valley event detection results of the compensated acceleration signal are used to construct the gait cycle, and the gait cycle is initially classified to obtain the original gait type; Based on information about stable human movement, the original gait type is smoothed in stages to obtain the smoothed human gait type. The current gait state is determined and counted based on the smoothed human gait type, thereby realizing the recognition of the human gait state.

2. The gait state recognition method based on acceleration signals according to claim 1, characterized in that, The step of acquiring the target axis acceleration signal at the waist position of the human body and performing signal preprocessing to obtain the compensated acceleration signal specifically includes: Based on preset frequency information, the target axis acceleration signal at the waist position of the human body is continuously sampled by an accelerometer to obtain the original acceleration sequence signal. The original acceleration sequence signal is processed to unify the direction, resulting in a unified original acceleration sequence signal. Based on the preset baseline estimation window length, baseline estimation is performed on the unified original acceleration sequence signal to obtain the baseline value; The difference between the unified original acceleration sequence signal and the baseline value is obtained to obtain the compensated acceleration signal.

3. The gait state recognition method based on acceleration signals according to claim 2, characterized in that, The step of detecting the current dynamic and static state of the human body based on the compensated acceleration signal and determining the stable motion state information of the human body specifically includes: Based on the first preset sliding window, the mean value of the compensated acceleration signal is calculated; Based on the first preset sliding window, and combined with the mean of the compensated acceleration signal, the variance of the compensated acceleration signal is calculated. Based on the window variance distribution of the compensated acceleration signals under preset human static state and walking motion state, the preset motion threshold is determined; The variance of the compensated acceleration signal is compared with a preset motion threshold to determine the current state of the human body, which includes the original motion state and the original stationary state. Based on the second preset sliding window, the current state of the human body is proportionally determined to determine the proportion of human body movement. Combined with the preset movement determination proportion threshold, the stable movement state information of the human body is determined. The stable movement state information of the human body includes the movement initiation stage and the stable movement stage.

4. The gait state recognition method based on acceleration signals according to claim 3, characterized in that, The step of performing peak-valley event detection on the compensated acceleration signal, determining whether the compensated acceleration signal is a peak event or a valley event, and outputting the peak-valley event detection result of the compensated acceleration signal specifically includes: Based on the third preset sliding window, the window mean of the compensated acceleration signal is calculated; Based on the third preset sliding window, and combined with the window mean of the compensated acceleration signal, the average absolute deviation of the compensated acceleration signal is calculated. Based on the window mean of the compensated acceleration signal and the mean absolute deviation of the compensated acceleration signal, an adaptive peak threshold and an adaptive valley threshold are constructed. If the compensated acceleration signal of the current sampling point is greater than the compensated acceleration signals of its adjacent sampling points, and the compensated acceleration signal of the current sampling point is greater than the adaptive peak threshold, then the compensated acceleration signal of the current sampling point is marked as a peak event. If the compensated acceleration signal of the current sampling point is less than the compensated acceleration signals of its adjacent sampling points, and the compensated acceleration signal of the current sampling point is less than the adaptive valley threshold, then the compensated acceleration signal of the current sampling point is marked as a valley event.

5. The gait state recognition method based on acceleration signals according to claim 4, characterized in that, It also includes performing fallback confirmation on the compensated acceleration signal of the current sampling point marked as a peak event, calculating the peak candidate amplitude of the compensated acceleration signal of the current sampling point, and calculating the difference between the compensated acceleration signal and the adjacent previous and next sampling points. When the difference is greater than the fallback confirmation coefficient multiplied by the dynamic noise estimate of the compensated acceleration signal of the current sampling point, it is determined that the fallback confirmation condition is met, and the compensated acceleration signal of the current sampling point is confirmed as a peak event. If the fallback confirmation condition is not met, the peak event is removed.

6. The gait state recognition method based on acceleration signals according to claim 5, characterized in that, The step of using the peak and valley event detection results of the compensated acceleration signal to construct gait cycles and perform initial classification of gait cycles to obtain the original gait type specifically includes: The gait period is constructed by combining the valley event of the compensated acceleration signal at the previous sampling point, the peak event of the compensated acceleration signal at the current sampling point, and the valley event of the compensated acceleration signal at the next sampling point. The rising and falling time-series characteristics of the gait cycle are determined. The rising time-series characteristics are determined by the difference between the peak event of the compensated acceleration signal at the current sampling point and the valley event of the compensated acceleration signal at the previous sampling point. The falling time-series characteristics are determined by the difference between the valley event of the compensated acceleration signal at the next sampling point and the peak event of the compensated acceleration signal at the current sampling point. The characteristic ratio of the gait cycle is determined based on the ratio between the rising and falling time-series characteristics of the gait cycle. The current gait cycle is initially classified based on the characteristic ratio of the gait cycle to obtain the original gait type.

7. The gait state recognition method based on acceleration signals according to claim 6, characterized in that, The step of initially classifying the current gait cycle based on the feature ratio of the gait cycle to obtain the original gait type specifically includes: If the feature ratio of the current gait cycle is greater than the preset forward gait determination threshold, then the original gait type of the current gait cycle is determined to be a forward state. If the feature ratio of the current gait cycle falls within the preset backward gait determination threshold range, then the original gait type of the current gait cycle is determined to be backward gait. If the absolute value of the difference between the feature ratio of the current gait cycle and the feature ratio of the previous valid gait cycle is less than the preset tolerance for stationary gait determination or does not satisfy the forward and backward gait conditions, then the original gait type of the current gait cycle is determined to be stationary gait.

8. The gait state recognition method based on acceleration signals according to claim 7, characterized in that, The step of performing a phased smoothing process on the original gait type based on human stable motion state information to obtain a smoothed human gait type specifically includes: If the stable motion state information of the human body is the initial stage of motion, start voting is performed on multiple consecutive original gait types to obtain the gait type of the initial stage. If the human body's stable motion state information is in the stable motion stage, a steady-state vote is performed on several original gait types to obtain a stable output gait type; By combining the gait type at the initial stage with the gait type at stable output, a smoothed human gait type is constructed.

9. The gait state recognition method based on acceleration signals according to claim 8, characterized in that, The step of determining the current gait state based on the smoothed human gait type and counting it to achieve gait state recognition specifically includes: The smoothed human gait types are voted on to obtain the current gait state; For the stable output gait type in the smoothed human gait type, the total number of steps, forward steps, backward steps and stationary steps are accumulated to obtain the current movement step count statistics; By combining the current gait status with the current number of steps, the gait status of the human body can be identified.

10. A gait state recognition system based on acceleration signals, characterized in that, Includes the following modules: The first module is used to collect the target axis acceleration signal at the waist position of the human body and perform signal preprocessing to obtain the compensated acceleration signal. The second module is used to detect the current dynamic and static state of the human body based on the compensated acceleration signal and determine the stable motion state information of the human body. The third module is used to perform peak and valley event detection on the compensated acceleration signal, determine whether the compensated acceleration signal is a peak event or a valley event, and output the peak and valley event detection results of the compensated acceleration signal. The fourth module is used to detect peak and valley events in the compensated acceleration signal, construct gait cycles, and perform initial classification of gait cycles to obtain the original gait type. The fifth module is used to perform phased smoothing of the original gait type based on the human body's stable motion state information, so as to obtain the smoothed human gait type. The sixth module is used to determine the current gait state based on the smoothed human gait type and count it, thereby realizing the recognition of the human gait state.

Images