A wristband fit detection method based on PPG signal

Abstract

The present invention relates to the field of data processing technology and discloses a wristband fit detection method based on PPG signals, comprising the following steps: step 1: acquiring a wristband PPG signal; step 2: performing fluctuation detection on three-channel data of the PPG signal respectively to obtain fluctuation statistics within a fixed period; step 3: weighting the statistical results obtained in step 2 to obtain a fluctuation probability value; step 4: obtaining a required judgment result based on the fluctuation probability value obtained in step 3 and a preset threshold. The method of the present invention can effectively distinguish waveform changes in tight and loose conditions through the PPG signal, and has a high detection accuracy.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a wristband fit detection method based on PPG signals. Background Art

[0002] Wearable devices, such as smart bracelets, have become an integral part of our lives, used to track daily activities, sleep patterns, and dietary habits. Wearable devices typically monitor users' health, exercise, and other data in real time or over time to understand their physical condition. Photoelectric sensors detect PPG data and analyze it to derive physiological parameters such as heart rate and blood oxygen saturation. A loose fit on the bracelet can increase noise in the collected PPG signal, affecting the accuracy of physiological signal calculations. The bracelet can also stop signal collection based on the resulting determination to conserve battery power.

[0003] The mainstream technology used in wearable bracelets on the market is the photoplethysmography (PPG) method. The principle is to use the light absorption ability of different tissues in the human body to obtain blood flow information and then obtain some physiological information of the human body. It is mainly divided into transmission type and reflection type. Due to the convenience of structural design, the same-side reflection type is generally used.

[0004] How to use the wearable device's own sensors to determine the wearing status is an urgent problem to be solved. Summary of the Invention

[0005] In order to solve the problems existing in the prior art, the present invention provides a wristband fit detection method based on PPG signals.

[0006] The technical solution adopted by the present invention is: a wristband fit detection method based on PPG signals, comprising the following steps:

[0007] Step 1: Get the PPG signal from the wristband;

[0008] Step 2: Perform fluctuation detection on the three-channel data of the PPG signal respectively to obtain the fluctuation statistics within a fixed period;

[0009] Step 3: Weight the statistical results obtained in step 2 to obtain the fluctuation probability value;

[0010] Step 4: Obtain the required judgment result based on the fluctuation probability value obtained in step 3 and the preset threshold.

[0011] Furthermore, the fluctuation detection process in step 2 is as follows:

[0012] The sliding window is used for detection to obtain the data change sequence;

[0013] Derivative the data change sequence to obtain the data change sequence;

[0014] The data fluctuation value is obtained by taking the difference of the data change sequence;

[0015] The data fluctuation value is compared with the fluctuation threshold. If it is within the threshold, it is determined to be fluctuation.

[0016] Furthermore, the fluctuation probability value in step 4 is compared with a preset threshold value. If it is within the preset threshold value, it is determined to be a tight fit; if it exceeds the threshold value, it is determined to be loose.

[0017] Furthermore, the weighting process in step 3 is as follows:

[0018]

[0019] in, Y is the weighted result value, A 、 B 、 C are the weighted values ​​corresponding to the red light signal, infrared light signal and green light signal respectively; RED is the fluctuation statistical result of the red light signal within a fixed period; IRED is the statistical result of the fluctuation of the near-infrared light signal within a fixed period, GREEN It is the statistical result of green light fluctuation in a fixed period.

[0020] Furthermore, the fluctuation threshold is determined as follows:

[0021]

[0022] in, is the fluctuation threshold, K 1. K 2 and K 3 is the difference range when fluctuation occurs under different wearing conditions, a is the preset coefficient value.

[0023] Furthermore, the calculation method of the difference range when fluctuations occur in different wearing states is as follows:

[0024]

[0025]

[0026]

[0027] in, V 1. V 2. V 3 are the output values ​​in static and tight wearing, slipping, and slightly loose wearing without slipping;

[0028] in, , ,

[0029] in, d 1. d 2. d 3 are the PPG sensor output values ​​in the corresponding states; N is the total number of outputs.

[0030] Furthermore, the weighted values ​​corresponding to the red light signal, infrared light signal, and green light signal are calculated as follows:

[0031]

[0032] in, M is the corresponding weighted value, j =1 A , j =2 B , j =3 C , is the signal-to-noise ratio of a single-channel signal, is the sum of the signal-to-noise ratios of the three signals, λ j is the wavelength, N j is the total number of data points for each signal, is the sampling rate of a single-channel signal, b is the proportionality coefficient, is the spectral volume pulse wave amplitude.

[0033] Furthermore, the spectrum volume pulse wave amplitude calculation method is as follows:

[0034]

[0035] in, is the total reflection coefficient corresponding to the wavelength in the diastolic state, is the total reflection coefficient corresponding to the wavelength in the contracted state.

[0036] The beneficial effects of the present invention are:

[0037] (1) The method of the present invention can effectively distinguish between loose and tight wearing of PPG devices through changes in PPG signal waveforms;

[0038] (2) The method of the present invention has been tested and has a high accuracy rate. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 Schematic diagram of the process of the present invention.

[0040] Figure 2 Schematic diagram of the fluctuation detection process in the present invention.

[0041] Figure 3 These are the PPG three-channel signal images collected under different wearing conditions, a is red light, b is infrared light, and c is green light.

[0042] Figure 4 For Figure 3 Three-axis signal diagram collected under the same conditions, a is the X-axis, b is the Y-axis, and c is the Z-axis.

[0043] Figure 5 This is a diagram showing an example of fluctuation detection in an embodiment of the present invention.

[0044] Figure 6 Schematic diagram of the relationship between the spectral volume pulse wave amplitude and wavelength in an embodiment of the present invention, a is the change of the total reflectance corresponding to the wavelength in the diastolic state with wavelength, and b is the spectral volume pulse wave amplitude value.

[0045] Figure 7 Schematic diagram of test accuracy statistical results in an embodiment of the present invention. DETAILED DESCRIPTION

[0046] The present invention will be further described below with reference to the accompanying drawings and specific embodiments.

[0047] like Figure 1 As shown, a wristband fit detection method based on PPG signals includes the following steps:

[0048] Step 1: Get the PPG signal from the wristband;

[0049] The collected PPG signal is three-channel data (Red, IRed, Green) with a sampling rate of 50 Hz. The length of the sliding window varies according to the sampling rate, and each channel will judge the signal fluctuation during the Tjudge period.

[0050] Step 2: Perform fluctuation detection on the three-channel data of the PPG signal respectively to obtain the fluctuation statistics within a fixed period;

[0051] Fluctuation detection methods such as Figure 2 As shown,

[0052] The sliding window is used for detection to obtain the data change sequence;

[0053] Derivative the data change sequence to obtain the data change sequence;

[0054] The data fluctuation value is obtained by taking the difference of the data change sequence;

[0055] The data fluctuation value is compared with a preset threshold value. If it is within the threshold value, it is determined to be a fluctuation.

[0056] The fluctuation threshold is determined as follows:

[0057]

[0058] in, is the fluctuation threshold, K 1. K 2 and K 3 is the difference range when fluctuation occurs under different wearing conditions, a is the preset coefficient value.

[0059] The calculation method for the difference range when fluctuations occur in different wearing states is as follows:

[0060]

[0061]

[0062]

[0063] in, V 1. V 2. V 3 are the output values ​​in static and tight wearing, slipping, and slightly loose wearing without slipping;

[0064] in, , ,

[0065] in, d 1. d 2. d 3 are the PPG sensor output values ​​in the corresponding states; N is the total number of outputs.

[0066] Step 3: Weight the statistical results obtained in step 2 to obtain the fluctuation probability value;

[0067] The weighting process is as follows:

[0068]

[0069] in, Y is the weighted result value, A 、 B 、 C are the weighted values ​​corresponding to the red light signal, infrared light signal and green light signal respectively; RED is the fluctuation statistical result of the red light signal within a fixed period; IRED is the statistical result of the fluctuation of the near-infrared light signal within a fixed period, GREEN It is the statistical result of green light fluctuation in a fixed period.

[0070] The weighted values ​​corresponding to the red light signal, infrared light signal, and green light signal are calculated as follows:

[0071]

[0072] in, M is the corresponding weighted value, j =1 A , j =2 B , j =3 C , is the signal-to-noise ratio of a single-channel signal, is the sum of the signal-to-noise ratios of the three signals, λ j is the wavelength, N j is the total number of data points for each signal, is the sampling rate of a single-channel signal, b is the proportionality coefficient, is the spectral volume pulse wave amplitude.

[0073] The spectral volume pulse wave amplitude is calculated as follows:

[0074]

[0075] in, is the total reflection coefficient corresponding to the wavelength in the diastolic state, is the total reflection coefficient corresponding to the wavelength in the contracted state.

[0076] Step 4: Obtain the required judgment result based on the fluctuation probability value obtained in step 3 and the preset threshold.

[0077] The fluctuation probability value is compared with a preset threshold. If it is within the preset threshold, it is determined to be a tight fit; if it exceeds the threshold, it is determined to be loose.

[0078] Example

[0079] The present invention will be further described below with reference to specific embodiments.

[0080] A wristband fit detection method based on PPG signals includes the following steps:

[0081] Step 1: Get the PPG signal from the wristband;

[0082] The collected PPG signal is three-channel data (Red, IRed, Green) with a sampling rate of 50 Hz. The length of the sliding window varies according to the sampling rate, and each channel will judge the signal fluctuation during the Tjudge period.

[0083] Currently, PPG sensors and three-axis accelerometers are usually used to collect corresponding signals. Figure 3 and Figure 4 The collected PPG three-channel signal and three-axis signal diagram are given. Figure 4 As can be seen from the boxed part bc, the waveform changes of the triaxial accelerometer when worn loosely and tightly are consistent in some cases. It is not possible to judge whether the wear is tight based on the sudden change of the triaxial accelerometer waveform. Figure 3 The PPG waveform shows obvious differences in different wearing scenarios, so the PPG signal is used as the data source.

[0084] Step 2: Perform fluctuation detection on the three-channel data of the PPG signal respectively to obtain the fluctuation statistics within a fixed period;

[0085] In this embodiment, the fluctuation detection example is as follows: Figure 5 As shown:

[0086] Use sliding window to detect and obtain data change sequence; confirm the length of sliding window data:

[0087] Based on the frequency of the signals sampled by the designed wristband device and the sudden changes in tightness and looseness during actual use, we determined the following: the sampling rate of the device is 50 Hz, and each data value corresponds to 20 ms, so five data values ​​correspond to 100 ms. Therefore, the length of each signal update interval is 5, and the selected sliding window length is 3.

[0088] The buffer is updated every time it is full of five data. The five data in the buffer array are slid three times, and the data changes with a window length of 3 are obtained each time. Finally, the sequence { y 1, y 2, y 3}, the number of criteria for judging whether the data set fluctuates changes from 5 to 3.

[0089] Derivative the data change sequence to obtain the data change sequence;

[0090] For the sequence { y 1, y 2, y 3} Take the first-order derivative and get the change sequence { y 1, y 2, y 3}Data changes { z1, z 2}, where z 1 represents the data change from Window 1 to Window 2. z 2 represents the data change from Window 2 to Window 3. At this time, the number of criteria for determining whether the data in the data window has changed is changed from 3 to 2.

[0091] The data fluctuation value is obtained by taking the difference of the data change sequence;

[0092] right{ z 1, z 2} Calculate the difference. When the difference is greater than a certain threshold, we can conclude that these five data have undergone data fluctuations corresponding to the situation where the bracelet is worn loosely.

[0093] The data fluctuation value is compared with the preset threshold. If it is within the threshold, it is determined to be a fluctuation. If it is a fluctuation, the output record is 1, and if it is not a fluctuation, the output is 0. The output results are shown below.

[0094] Red: {0, 1, 0, 1, 1, 1, 1, 0, 1, 1, 1,…………}

[0095] IRed: {0, 0, 1, 1, 1, 1, 1, 0, 1, 0, 1,…………}

[0096] Green: {1, 1, 1, 0, 1, 0, 0, 0, 1, 1, 1,…………}

[0097] The total number of single-channel outputs is N

[0098] The fluctuation threshold is determined as follows:

[0099]

[0100] in, is the fluctuation threshold, K 1. K 2 and K 3 is the difference range when fluctuation occurs under different wearing conditions, a is the preset coefficient value. In this embodiment, it is 7 / 50. This value is based on the degree of change of the waveform combined with K 1. K 2 and K 3. Obtained through comprehensive judgment.

[0101] The calculation method for the difference range when fluctuations occur in different wearing states is as follows:

[0102]

[0103]

[0104]

[0105] in, V 1. V 2. V 3 are the output values ​​in the static and tight wearing state, the state with sliding, and the state with slightly loose wearing but no sliding; the above three values ​​are the PPG output values ​​obtained by the detection of obvious tight wearing, sliding, and slight sliding.

[0106] in, , ,

[0107] in, d 1. d 2. d 3 are the PPG sensor output values ​​in the corresponding states; N is the total number of outputs.

[0108] Step 3: Weight the statistical results obtained in step 2 to obtain the fluctuation probability value;

[0109] The weighting process is as follows:

[0110]

[0111] in, Y is the weighted result value, A 、 B 、 C are the weighted values ​​corresponding to the red light signal, infrared light signal and green light signal respectively; RED is the fluctuation statistical result of the red light signal within a fixed period; IRED is the statistical result of the fluctuation of the near-infrared light signal within a fixed period, GREEN It is the statistical result of green light fluctuation in a fixed period.

[0112] in , , ;

[0113] n 1. n 2. n 3 are the number of fluctuations counted in step 2 for red light, infrared light and green light respectively.

[0114] The weighted values ​​corresponding to the red light signal, infrared light signal, and green light signal are calculated as follows:

[0115]

[0116] in, M is the corresponding weighted value, j=1 A , j =2 B , j =3 C , is the signal-to-noise ratio of a single-channel signal, is the sum of the signal-to-noise ratios of the three signals, λ j is the wavelength, N j is the total number of data points for each signal, is the sampling rate of a single-channel signal, b is the proportionality coefficient, is the spectral volume pulse wave amplitude, the curve is as follows Figure 6 As shown in b.

[0117] The spectral volume pulse wave amplitude is calculated as follows:

[0118]

[0119] in, is the total reflection coefficient corresponding to the wavelength in the diastolic state, and its curve is as follows Figure 6 As shown in a, is the total reflection coefficient corresponding to the wavelength in the contracted state.

[0120] Step 4: Obtain the required judgment result based on the fluctuation probability value obtained in step 3 and the preset threshold.

[0121] The fluctuation probability value is compared with a preset threshold. If it is within the preset threshold, it is determined to be a tight fit. If it exceeds the threshold, it is determined to be loose. For example, in this embodiment, the preset threshold is 50%. If it exceeds 50%, it is determined to be loose. Otherwise, it is determined to be a tight fit.

[0122] To illustrate the effectiveness of the method, five volunteers of different genders and wrist circumferences were tested using the method to simulate sports scenarios by swinging their arms back and forth and quickly turning their wrists 90 degrees. All 100 tests with tight fit resulted in accurate judgments, with an accuracy rate of 100%. However, the five volunteers made only one incorrect judgment with loose fit, resulting in an accuracy rate of 99% for all 100 tests with loose fit. The statistical results are as follows: Figure 7 shown.

[0123] Existing wristbands utilize the three-channel signal extracted from PPG, but they only represent physiological information and do not further process the signal's fluctuations. The method of the present invention, by processing the three-channel signal, can accurately determine the wearing condition of a wristband device with high accuracy.

Claims

1. A wristband fit detection method based on PPG signals, characterized in that: The following steps are involved: Step 1: Get the PPG signal from the wristband; Step 2: Perform fluctuation detection on the three-channel data of the PPG signal respectively to obtain fluctuation statistics within a fixed period, wherein the three-channel data includes red light signal, infrared light signal and green light signal; The fluctuation detection process is as follows: The sliding window is used for detection to obtain the data change sequence; Derivative the data change sequence to obtain the data change sequence; The data fluctuation value is obtained by taking the difference of the data change sequence; The data fluctuation value is compared with the fluctuation threshold. If it is within the threshold, it is determined to be a fluctuation; Step 3: Weight the statistical results obtained in step 2 based on the spectral volume pulse wave amplitude to obtain a fluctuation probability value; The weighted values ​​corresponding to the red light signal, infrared light signal, and green light signal are calculated as follows: in, M is the corresponding weighted value, j =1 A , j =2 B , j =3 C , is the signal-to-noise ratio of a single-channel signal, is the sum of the signal-to-noise ratios of the three signals, λ is the wavelength, N j is the total number of data points for each signal, is the sampling rate of a single-channel signal, b is the proportionality coefficient, is the spectral volume pulse wave amplitude; The spectrum volume pulse wave amplitude calculation method is as follows: in, is the total reflection coefficient corresponding to the wavelength in the diastolic state, is the total reflection coefficient corresponding to the wavelength in the contracted state; Step 4: Obtain the required judgment result based on the fluctuation probability value obtained in step 3 and the preset threshold.

2. The wristband fit detection method based on PPG signal according to claim 1, characterized in that: In step 4, the fluctuation probability value is compared with a preset threshold. If the value is within the preset threshold, it is determined to be a tight fit; if the value exceeds the threshold, it is determined to be loose.

3. The wristband fit detection method based on PPG signal according to claim 1, characterized in that: The weighting process in step 3 is as follows: in, Y is the weighted result value, A 、 B 、 C are the weighted values ​​corresponding to the red light signal, infrared light signal and green light signal respectively; RED is the fluctuation statistical result of the red light signal within a fixed period; IRED is the statistical result of the fluctuation of the near-infrared light signal within a fixed period, GREEN It is the statistical result of green light fluctuation in a fixed period.

4. The wristband fit detection method based on PPG signals according to claim 1, characterized in that: The fluctuation threshold determination method is as follows: in, is the fluctuation threshold, K 1. K 2 and K 3 is the difference range when fluctuation occurs under different wearing conditions, a is the preset coefficient value.

5. The wristband fit detection method based on PPG signals according to claim 4, characterized in that: The calculation method of the difference range when fluctuations occur in different wearing states is as follows: in, V 1. V 2. V 3 are the output values ​​in static and tight wearing, slipping, and slightly loose wearing without slipping; in, , , in, d 1. d 2. d are the PPG sensor output values ​​in the corresponding states respectively; N is the total number of outputs.

Images