Stochastic resonance enhanced smartphone imu acceleration data step detection method

Abstract

A kind of random resonance enhanced smart phone IMU acceleration data step detection method, using the initialization of a group of acceleration data of known step number, random resonance parameter and step detection threshold;According to the initialization of random resonance parameter, the acceleration data to be measured is processed by bistable random resonance model, and the processed acceleration data is detected according to the initialization of step detection threshold, to obtain the final step detection result.The acceleration data of the present application is processed using random resonance, the peak uniformity of the original acceleration data is improved, and under the condition that the detection parameter is fixed, compared with directly using the original acceleration data for step detection, the acceleration data processed by random resonance significantly improves the peak uniformity of acceleration waveform, reduces the influence of noise, and improves the step detection rate.

Description

Technical Field

[0001] This invention relates to a technology in the field of gait detection, specifically a gait detection method based on random resonance-enhanced smartphone IMU acceleration data. Background Technology

[0002] In pedestrian dead reckoning (PDR) positioning, acceleration data is commonly used for step detection. However, due to cost limitations, the acceleration measurement data from smartphones' built-in inertial measurement units (IMUs) has poor peak uniformity and exhibits spikes, which is detrimental to step detection. Stochastic resonance is used to enhance acceleration data, improve the peak consistency of acceleration data, and increase the probability of step detection. Summary of the Invention

[0003] To address the issues of poor uniformity and glitches in existing smartphone-embedded IMU acceleration data, a stochastic resonance-enhanced method for step detection using smartphone IMU acceleration data is proposed. This method uses stochastic resonance to process the smartphone IMU acceleration data, thereby improving the peak uniformity of the original acceleration data. Furthermore, with fixed detection parameters, the acceleration data processed by stochastic resonance significantly improves the peak uniformity of the acceleration waveform compared to using the original acceleration data directly for step detection, reduces the impact of noise, and increases the step detection rate.

[0004] This invention is achieved through the following technical solution:

[0005] This invention relates to a stochastic resonance-enhanced method for step detection of acceleration data from a smartphone IMU. The method initializes stochastic resonance parameters and a step detection threshold using a set of acceleration data with a known number of steps. Based on the initialized stochastic resonance parameters, the acceleration data to be measured is processed through a bistable stochastic resonance model. Step detection is then performed on the processed acceleration data based on the initialized step detection threshold to obtain the final step detection result.

[0006] The aforementioned bistable stochastic resonance model refers to the following: Brownian particles, driven by weak periodic forces and random forces within a bistable potential well, repeatedly jump within the system's potential well. Driven by periodic signals and Gaussian white noise, the Langevin equation, possessing the properties of a double potential well, is a typical model describing stochastic resonance systems. Specifically: Where: s(t) is the weak periodic signal input to the bistable system, i.e., the input acceleration data; x(t) is the output signal of the bistable system, i.e., the acceleration data after processing by the stochastic resonance system; n(t) represents the noise term; V' represents the derivative operation with respect to V; and the stochastic resonance potential function is... in: These are the parameters of the bistable situation trap. By adjusting the system parameters a and b, bistable situation functions of different shapes can be obtained.

[0007] The initialization of the random resonance parameters and the step detection threshold refers to: based on a set of known step counts S true Acceleration data A ref (t) is used as the input to the stochastic resonance system, i.e., s(t) = A ref The output x(t) is obtained by adjusting parameters a and b. The waveform shape of x(t) is judged by comparing the waveform states corresponding to different parameters a and b, and the optimal waveform shape x is selected. opt The values ​​of a' and b' corresponding to (t) are the initialized random resonance parameters; Step detection is performed on the acceleration data after random resonance processing using a' and b': S = ∑I[x(t)≥Th], where: S represents the number of step detections; Σ represents the summation operation; I[*] is an indicator function, which takes 1 if the input value ★ is true, and 0 if ★ is false; Th represents the threshold for step detection on the acceleration data; x opt (t) is used as input, i.e., x(t) = x opt (t) and S is obtained by adjusting the parameter Th. true =S, and the initial step detection threshold Th' can be obtained.

[0008] The processing of the acceleration data to be measured refers to: based on the stochastic resonance parameters a', b' and the acceleration data to be measured A Ori (After substituting t into the bistable stochastic resonance model, the acceleration data after stochastic resonance processing is obtained by using the fourth-order Runge-Kutta function.)

[0009] The step detection of the processed acceleration data refers to substituting the initialized step detection threshold and the acceleration data after random resonance processing into S = ∑I[x(t)≥Th] to perform step detection, and obtaining the step detection result S. SR .

[0010] This invention relates to a smartphone IMU acceleration data step detection system for implementing the above method, comprising: a parameter initialization module, a random resonance processing module, and a step detection module, wherein: the parameter initialization module initializes random resonance parameters and a step detection threshold based on a set of acceleration data with a known number of steps; the random resonance processing module processes the input acceleration data to be measured using the random resonance parameters obtained by the parameter initialization module; and the step detection module applies the step detection threshold to detect the acceleration data processed by the random resonance processing module to obtain a step detection result. Attached Figure Description

[0011] Figure 1 This is a flowchart of the present invention;

[0012] Figure 2 Acceleration data from a smartphone IMU with a known number of steps, used for initialization;

[0013] Figure 3 This is a graph showing the results of processing a set of smartphone IMU acceleration data during a slow walking test.

[0014] Figure 4 This is a graph showing the results of processing a set of smartphone IMU acceleration data during a fast walking test.

[0015] Figure 5 This is a graph showing the results of processing a set of smartphone IMU acceleration data during a running test. Detailed Implementation

[0016] This embodiment uses an OPPO REALME RMX3366 smartphone running Android 13 to collect acceleration data. It has a built-in Bosch BMI26X accelerometer sensor with an acceleration sampling frequency of 51Hz. A volunteer holds the phone and collects 100 steps of IMU data while walking. This data is used as acceleration data for parameter initialization based on known steps. Then, 100 steps of data are collected each in three different states: slow walking, fast walking, and running, for testing. The acquired data is post-processed using MATLAB 2023a software. The specific steps include:

[0017] Step 1: Initialize the parameters using a set of acceleration data with a known number of steps, specifically including:

[0018] 1.1) Given a set of steps with a known number of steps of 100, i.e., S true =100 raw acceleration data A ref (t) is input into the parameter initialization module, A ref The graph of (t) is as follows Figure 2 As shown in (a);

[0019] 1.2) Input A ref (t) is used as the input to the stochastic resonance system, i.e., s(t) = A ref (t), adjusting parameters a and b yields the output x(t). The waveform shape of x(t) is judged, and the waveform states corresponding to different parameters a and b are compared. A' = 1 and b' = 1000 are selected as the optimal stochastic resonance parameters, and the corresponding output waveform x is obtained. opt (t) such as Figure 2 As shown in (b).

[0020] 1.3) x opt Using x(t) as x(t), the value of Th is continuously adjusted until S = 100, resulting in an initial step detection threshold Th' = 0.8. For comparison, the original acceleration data A is used... ref Substituting x(t) into the equation, we similarly adjust the value of Th so that S = 100, thus obtaining the detection threshold Th for gait detection using the raw acceleration data. Ori =0.8613.

[0021] The random resonance parameters and step detection thresholds obtained in steps 1.2 and 1.3 are summarized and displayed in Table 1. Figure 2 The red *marked detection points and magenta numbers in the table represent the gait detection results obtained by applying the gait detection thresholds in Table 1. The values ​​are all 100, indicating that the detection thresholds are effective. Figure 2 (b) The acceleration data after random resonance processing is shown in comparison. Figure 2 The original acceleration data in (a) has better peak consistency.

[0022] Table 1 shows the random resonance parameters and step detection thresholds obtained by the parameter initialization module.

[0023] project value Stochastic resonance parameters a = 1, b = 1000 Pace detection threshold (used for pace detection in raw acceleration data) 0.8613 Pace detection threshold (used for pacing detection in acceleration data after random resonance processing) 0.8

[0024] Step 2: Process the acceleration data acquired using a smartphone using stochastic resonance, specifically including:

[0025] To test the effectiveness of the method proposed in this invention, volunteers used a test phone to collect three sets of test data at different walking speeds, labeled A and B respectively. Ori ,1(t),A Ori ,2(t),A Ori 3(t), the three sets of data correspond to respectively Figures 3-5 Figure (a) in the middle.

[0026] 2.1) Combine the random resonance parameters a' = 1, b' = 1000 obtained in step 1 with the acceleration data to be measured, A Ori ,1(t),A Ori ,2(t),A Ori 3(t) are respectively input into the stochastic resonance processing module, that is, a'=1, b'=1000 as a, b, A Ori ,1(t),A Ori ,2(t),A Ori 3(t) is substituted into the bistable stochastic resonance model as s(t).

[0027] 2.2) The bistable stochastic resonance model is solved using the fourth-order Runge-Kutta function to obtain the acceleration data x after stochastic resonance processing.SR，1 (t), x SR，2 (t), x SR，3 (t), the three sets of data waveforms correspond to respectively Figures 3-5 Figure (b) in the middle.

[0028] Step 3: Perform gait detection on the acceleration data after random resonance processing. Specifically, this includes: using the gait detection threshold Tj' = 0.8 obtained in Step 1 and the acceleration data x obtained in Step 2 after random resonance processing. SR，1 (t), x SR，2 (t), x SR，3 x(t) is input as x(t) into the step detection module, Th = Th' = 0.8, and step detection is performed to obtain the step detection result S. SR，1 =100, S SR，2 =100, S SR，3 =101, respectively displayed in Figures 3-5 In Figure (b).

[0029] For comparison, the original acceleration data A Ori，1 (t), A Ori，2 (t), A Ori，3 (t) is x(t), and the step detection threshold Th Ori =0.8613 is input into the gait detection module to obtain the gait detection result S. Ori，1 =33, S Ori，2 =56, S Ori，3 =5, respectively displayed in Figures 3-5 In Figure (a).

[0030] contrast Figures 3-5 As shown in Figures (a) and (b), directly using the original acceleration for step detection yielded very poor results: 33, 56, and 5 steps detected, respectively. The third set of data, in particular, only detected 5 valid steps, while the correct number should be 100. In contrast, the detection results obtained using this invention are excellent, with 100, 100, and 101 steps detected. The first two sets show a step detection count consistent with the actual 100 steps, and the third set, at 101, is very close to the true value of 100. Therefore, the experimental results demonstrate that the acceleration data from the smartphone IMU enhanced by stochastic resonance is more stable, and with a fixed detection threshold, a higher step detection rate can be achieved.

[0031] The evaluation of the step detection effect is as follows: a detection threshold is obtained by applying a set of acceleration data with a known number of steps. The same threshold is used to perform step detection on other acceleration data to be tested collected by the mobile phone. The detection result is then compared with the actual number of steps. The detection effect is better if it is closer to the actual number of steps.

[0032] Compared with existing technologies, the smartphone IMU acceleration data is more stable and has a higher probability of step detection after random resonance processing.

[0033] The above-described specific implementations can be partially adjusted by those skilled in the art in different ways without departing from the principles and purpose of the present invention. The scope of protection of the present invention is defined by the claims and is not limited to the above-described specific implementations. All implementation schemes within the scope of the claims are bound by the present invention.

Claims

1. A method for step detection using stochastic resonance-enhanced smartphone IMU acceleration data, characterized in that, The random resonance parameters and step detection threshold are initialized using a set of acceleration data with a known number of steps. The acceleration data to be measured is processed by a bistable random resonance model based on the initialized random resonance parameters. Step detection is then performed on the processed acceleration data based on the initialized step detection threshold to obtain the final step detection result. The initialization of the random resonance parameters and the step detection threshold refers to: based on a set of known step counts... acceleration data As input to a stochastic resonance system, i.e. By adjusting parameters Get output ,right The waveform shape is used to determine the value, and different parameters are compared. Choose the optimal waveform shape based on the corresponding waveform state. corresponding The value is the initialized random resonance parameter; for use Step detection is performed on acceleration data after random resonance processing. ,in: Indicates the number of steps detected; This represents the summation operation; It is an indicator function; if the input value... If true, take 1; if If the result is false, then take 0; This represents the threshold for gait detection using acceleration data; As input, i.e. Furthermore, by adjusting the parameters get This allows us to obtain the initial step detection threshold. ; The adjustment parameters This refers to: stochastic resonance potential function ,in: For bistable state trap parameters, by adjusting system parameters The adjustment is used to obtain bistable state functions of different shapes; The processing of the acceleration data to be measured refers to: based on stochastic resonance parameters and the acceleration data to be measured After substituting into the bistable stochastic resonance model, the acceleration data after stochastic resonance processing is obtained by using the fourth-order Runge-Kutta function. The aforementioned step detection on the processed acceleration data refers to substituting the initialized step detection threshold and the acceleration data processed by random resonance into... Perform gait detection and obtain gait detection results. .

2. The method for step detection of smartphone IMU acceleration data with stochastic resonance enhancement according to claim 1, characterized in that, The aforementioned bistable stochastic resonance model refers to the following: Brownian particles, driven by weak periodic forces and random forces within a bistable potential well, repeatedly jump within the system's potential well. Driven by periodic signals and Gaussian white noise, the Langevin equation, possessing the properties of a double potential well, is a typical model describing stochastic resonance systems. Specifically: ,in: The input signal is a weak periodic signal for the bistable system, namely, the input acceleration data. The output signal of the bistable system is the acceleration data after processing by the stochastic resonance system. Indicates the noise term; Indicates to Perform differentiation.

3. A smartphone IMU acceleration data gait detection system for implementing the stochastic resonance enhancement method of claim 1 or 2, comprising: The system comprises a parameter initialization module, a random resonance processing module, and a step detection module. The parameter initialization module initializes the random resonance parameters and the step detection threshold based on a set of acceleration data with a known number of steps. The random resonance processing module processes the input acceleration data to be measured using the random resonance parameters obtained from the parameter initialization module. The step detection module applies the step detection threshold to detect the acceleration data processed by the random resonance processing module to obtain the step detection result.

Images