Single-point real-time pulse dominant wave interval sequence extraction method and device

Abstract

The invention discloses a single-point real-time pulse dominant wave interval sequence extraction method and device, and relates to the technical field of two-dimensional pulse signal analysis. In order to solve the technical defects that in the prior art, challenges exist in the aspects of real-time extraction and accurate detection of pulse dominant wave interval sequences, and particularly, the application in wearable equipment and mobile terminal environments is obviously limited, the technical scheme provided by the invention comprises the following steps: acquiring a two-dimensional pulse signal to be processed; performing multi-byte data analysis on the two-dimensional pulse signal to obtain a data stream; noise reduction and baseline correction are conducted on the pulse signals, and standard pulse waves are output; converting the waveform into a square wave to obtain a high-low level signal; filtering noise pulses with insufficient high-level width; and calculating a time difference between adjacent effective pulse periods, and outputting a pulse dominant wave interval sequence. The method can be applied to cardiovascular health monitoring, intelligent health management, real-time analysis of wearable equipment and the like.

Description

Technical Field

[0001] It relates to the technical field of two-dimensional pulse signal analysis, and specifically relates to the extraction of the pulse main wave interval sequence in real time at a single point. Background Art

[0002] With the continuous development of modern medicine and biosensing technology, pulse signal detection plays an increasingly important role in the early detection of cardiovascular diseases and the field of personal health monitoring. Traditional pulse signal analysis usually relies on electrocardiogram (ECG) devices and high-precision sensors to obtain complete and stable physiological signals for subsequent analysis. However, medical-grade devices such as ECG are costly and bulky, and often need to be used in medical institutions or professional scenarios, which is not conducive to people's convenient daily health monitoring. To further lower the monitoring threshold, many studies have begun to explore pulse signals collected based on photoplethysmography (PPG) to obtain human physiological information such as heart rate and vascular elasticity, and apply them in personal health management through wearable devices or mobile terminals.

[0003] In the prior art, the pulse signals collected by PPG are widely used to estimate heart rate variability (HRV), judge blood oxygen saturation, and analyze indicators such as vascular elasticity. Some research results based on PPG have proposed improved filtering methods and recognition algorithms to deal with noise such as motion artifacts, environmental light interference, and baseline drift in daily life. However, these algorithms often still have the following deficiencies or limitations:

[0004] Weak real-time performance

[0005] Many detection schemes need to first collect sufficient pulse data and then process and analyze it offline or semi-offline, which is difficult to meet the demand for real-time and accurate detection of the pulse main wave interval (PMWI).

[0006] High device dependence

[0007] Traditional methods require a more professional or expensive acquisition system to ensure the high quality of the signal. The pulse signals collected by ordinary wearable devices or mobile terminals under light interference and motion states are often not stable enough, resulting in limited detection accuracy.

[0008] Complex algorithms and insufficient portability

[0009] Some schemes require a large number of floating-point operations in the filtering or waveform recognition stage, which is a heavy burden on embedded hardware platforms or low-power mobile terminal devices, and the transplantation cost is high, which is not conducive to popularization in a wider range of practical applications.

[0010] Prone to motion artifacts and baseline drift

[0011] In the daily use environment, people may be in a state of walking, running, etc. at any time, and changes in environmental temperature, finger position, and measurement posture will also introduce large baseline drifts and noise interferences. Without targeted filtering and algorithm optimization, these interferences will reduce the accuracy and stability of detection.

[0012] In summary, the existing technologies still face certain challenges in the real-time extraction and accurate detection of the pulse main wave interval sequence, especially in the applications of wearable devices and mobile environments, which are significantly limited. How to effectively suppress motion artifacts, environmental light, and baseline drift while ensuring real-time performance and accuracy, and enable the algorithm to be efficiently deployed on resource-limited device terminals has become a technical problem to be solved urgently. Summary of the Invention

[0013] To solve the challenges existing in the real-time extraction and accurate detection of the pulse main wave interval sequence in the existing technologies, especially the obvious technical defects in the applications of wearable devices and mobile environments, the technical solution provided by the present invention is as follows:

[0014] A method for extracting a single-point real-time pulse main wave interval sequence, comprising the following steps:

[0015] The step of obtaining a two-dimensional pulse signal to be processed by using an all-inclusive photoplethysmogram acquisition system;

[0016] The step of performing multi-byte data parsing on the two-dimensional pulse signal to obtain a pulse signal data stream for subsequent operations;

[0017] The step of performing second-order IIR filtering on the pulse signal data stream, and using a cascaded integer coefficient notch filtering and baseline drift collaborative elimination algorithm for real-time noise reduction and baseline correction, and outputting a standard pulse wave;

[0018] The step of setting an amplitude threshold for the standard pulse wave and converting the waveform into a square wave to obtain a high and low level signal;

[0019] The step of detecting the falling edge of the square wave signal to determine the end of each pulse cycle and filtering out noise pulses with insufficient high level width;

[0020] The step of calculating the time difference between adjacent effective pulse cycles and real-time outputting the pulse main wave interval sequence.

[0021] Furthermore, a preferred implementation manner is provided. The all-inclusive photoplethysmogram acquisition system adopts a finger clip structure, which can form a closed space between the sensor and the finger to reduce external light interference.

[0022] Further, a preferred embodiment is provided, in which the second-order IIR filtering preferentially suppresses high-frequency noise and glitch signals, providing a smoother input signal for subsequent integer-coefficient notch filtering.

[0023] Further, a preferred embodiment is provided, in which the baseline drift collaborative cancellation algorithm dynamically corrects the slow-changing trend of the pulse waveform through a sliding window method to ensure that the overall mean value of the pulse waveform is within a set range.

[0024] Further, a preferred embodiment is provided, in which during the quantization step, while magnifying the standard pulse wave by a multiple, a threshold value near zero is selected to more accurately distinguish the high-level section and the low-level section.

[0025] Further, a preferred embodiment is provided, in which during the period detection step, by judging the high-level width, when the width exceeds a preset threshold, it is determined as a valid pulse period to eliminate false pulses caused by motion artifacts or burst noise.

[0026] Based on the same inventive concept, the present invention also provides a single-point real-time device for extracting the pulse main wave interval sequence, including the following modules:

[0027] A module that uses an all-inclusive photoplethysmogram acquisition system to obtain the two-dimensional pulse signal to be processed;

[0028] A module that performs multi-byte data parsing on the two-dimensional pulse signal to obtain a pulse signal data stream for subsequent operations;

[0029] A module that performs second-order IIR filtering on the pulse signal data stream and uses cascaded integer-coefficient notch filtering and baseline drift collaborative cancellation algorithm for real-time noise reduction and baseline correction, and outputs a standard pulse wave;

[0030] A module that sets an amplitude threshold for the standard pulse wave and converts the waveform into a square wave to obtain high and low level signals;

[0031] A module that detects the falling edge of the square wave signal to determine the end of each pulse period and filters out noise pulses with insufficient high-level width;

[0032] A module that calculates the time difference between adjacent valid pulse periods and outputs the pulse main wave interval sequence in real time.

[0033] Based on the same inventive concept, the present invention also provides a computer storage medium for storing a computing program, and when the computer program is read by a computer, the computer executes the method described above.

[0034] Based on the same inventive concept, the present invention also provides a computer, including a processor and a storage medium. When the processor reads a computer program stored in the storage medium, the computer executes the method described above.

[0035] Based on the same inventive concept, the present invention also provides a computer program product. As a computer program, when the computer program is executed, the method described above is implemented.

[0036] Compared with the prior art, the beneficial effects of the technical solution provided by the present invention are as follows:

[0037] By adopting an all-inclusive photoplethysmography technology acquisition system, interference caused by ambient light and finger posture changes is reduced from the source. Compared with traditional partially wrapped and open finger clips, the signal integrity and stability are significantly improved, and artifacts in the motion scenario are also effectively suppressed.

[0038] Through a cascaded processing method combining second-order IIR filtering and notch filtering based on integer coefficient optimization, sufficient filtering accuracy can be ensured on a low-power hardware platform, and the burden of floating-point operations can be significantly reduced. Compared with existing filtering schemes that require high-performance hardware support, the present invention greatly improves the portability of the filtering algorithm in wearable devices and mobile terminals.

[0039] Through a sliding window and baseline drift collaborative elimination algorithm, slow-changing trend interference can be eliminated in real time, ensuring that the pulse signal will not have serious baseline drift due to factors such as finger temperature, slight movement, and sensor contact pressure during the acquisition process. It has more advantages in fast response and continuous signal smoothness compared with existing step-by-step filtering methods.

[0040] By converting the filtered signal into a square wave based on an amplitude threshold and then relying on the method of detecting the falling edge period, the problems of excessive offline period positioning delay and inaccurate positioning caused by peak interference are solved. Compared with the traditional main wave positioning method of post-analysis, both the detection time delay and accuracy are significantly improved.

[0041] By calculating the PP interval through the high and low level widths of the quantized pulse signal, invalid pseudo-peaks and noise segments can be effectively eliminated while ensuring real-time performance, greatly improving the recognition accuracy of the pulse main wave interval. Compared with the old scheme that only relies on a single heart rate estimation strategy, the new method can more flexibly adapt to various usage environments.

[0042] It can be applied to work such as cardiovascular health monitoring, intelligent health management, and real-time analysis of wearable devices. Description of the Drawings

[0043] Figure 1 It is a flowchart of a method for extracting a single-point real-time pulse main wave interval sequence;

[0044] Figure 2 This is the flowchart of the original signal data processing for this embodiment;

[0045] Figure 3 This is the flowchart of the algorithm for real-time extraction of the pulse main wave interval sequence based on the two-dimensional pulse signal

[0046] Figure 4 This is the schematic diagram of the system structure for real-time extraction of the pulse main wave interval sequence of the mobile pulse signal processing and algorithm;

[0047] Figure 5 This is the overall flowchart of the system for real-time extraction of the pulse main wave interval sequence at a single point;

[0048] Figure 6 This is the schematic diagram of the algorithm for real-time extraction of the pulse main wave interval sequence;

[0049] Figure 7 This is the schematic diagram of the real-time pulse main wave interval sequence;

[0050] Figure 8 This is the result of the main wave interval sequence. Specific implementation mode

[0051] To make the advantages and beneficial effects of the technical solution provided by the present invention more clearly manifested, the technical solution provided by the present invention will be further described in detail below with reference to the accompanying drawings. Specifically:

[0052] Embodiment 1. This embodiment provides a method for real-time extraction of the pulse main wave interval sequence at a single point, including the following steps:

[0053] The step of obtaining the two-dimensional pulse signal to be processed by using an all-inclusive photoplethysmogram acquisition system;

[0054] The step of performing multi-byte data parsing on the two-dimensional pulse signal to obtain a pulse signal data stream for subsequent operations;

[0055] The step of performing second-order IIR filtering on the pulse signal data stream and using a cascade of integer coefficient notch filtering and baseline drift collaborative elimination algorithm for real-time noise reduction and baseline correction, and outputting a standard pulse wave;

[0056] The step of setting an amplitude threshold for the standard pulse wave and converting the waveform into a square wave to obtain a high and low level signal;

[0057] The step of detecting the falling edge of the square wave signal to determine the end of each pulse cycle and filtering out noise pulses with insufficient high level width;

[0058] The step of calculating the time difference between adjacent effective pulse cycles and real-time outputting the pulse main wave interval sequence.

[0059] The described all - in - one photoplethysmogram acquisition system adopts a finger - clip structure, which can form a closed space between the sensor and the finger to reduce external light interference.

[0060] The second - order IIR filter preferentially suppresses high - frequency noise and glitch signals, providing a smoother input signal for subsequent integer - coefficient notch filtering.

[0061] The baseline drift collaborative elimination algorithm dynamically corrects the slow - changing trend of the pulse waveform through a sliding window method to ensure that the overall mean value of the pulse waveform is within a set interval.

[0062] The quantization step, while magnifying the standard pulse wave by a multiple, selects a threshold near zero to more accurately distinguish the high - level section and the low - level section.

[0063] The period detection step determines the effective pulse period by judging the width of the high - level pulse. When the width exceeds the preset threshold, it is determined as an effective pulse period to eliminate false pulses caused by motion artifacts or sudden noises.

[0064] Embodiment 2: This embodiment further explains in detail the technical solution provided in Embodiment 1. Specifically:

[0065] This embodiment provides a real - time extraction scheme for the pulse main - wave interval sequence based on a mobile device or a wearable device, including core processes such as signal acquisition, filtering processing, waveform quantization, and period detection. After obtaining the original pulse signal through the all - in - one photoplethysmogram technology acquisition system, the signal is pre - processed on an embedded device or a mobile terminal using a multiple - filtering algorithm. Finally, the filtered signal is converted into a square wave, and the time interval between adjacent pulse main - waves is calculated based on the falling edge of the square wave to achieve the real - time extraction and output of the pulse main - wave interval sequence.

[0066] Specific implementation steps

[0067] Step 1: Pulse signal acquisition and preliminary pre - processing

[0068] 1.1 Wear or attach the all - in - one photoplethysmogram technology acquisition system to the fingertip to ensure that the finger clip can isolate external light to the greatest extent.

[0069] 1.2 Continuously obtain the human pulse waveform signal at the sampling frequency of the acquisition device (which can be set to hundreds of times per second or higher according to requirements). At this time, the obtained original pulse signal values may be affected by factors such as ambient light changes, motion artifacts, and noise interference.

[0070] 1.3 Parse the collected multi - byte data into a floating - point or integer data stream for subsequent arithmetic processing to complete the preliminary arrangement of the original data.

[0071] 1.4 The original pulse signal data stream output by this step will be directly used as the input for the next filtering process.

[0072] Step 2: Multistage filtering process

[0073] 2.1 Based on the obtained pulse signal data stream, first use the second-order filtering algorithm to remove sharp spikes and high-frequency noise in the data. This filtering process can be implemented on low-power hardware or mobile devices, ensuring computational efficiency while removing most interference components.

[0074] 2.2 Input the preliminary cleaned signal output from the previous filtering process into the integer coefficient notch filtering and baseline drift co-cancellation module. By reasonably using integer operations and circular buffer management on embedded devices, the floating-point operation burden on the hardware can be significantly reduced, and synchronous cancellation of power frequency interference and slow baseline drift can be achieved in real-time scenarios.

[0075] 2.3 After the above two-stage filtering, a relatively standard and stable pulse waveform can be obtained, reducing the up and down fluctuations of the baseline and spike noise. This filtered signal will be used as the input for the next quantization and square wave conversion.

[0076] Step 3: Pulse waveform quantization and threshold processing

[0077] 3.1 According to the standard pulse waveform signal output in Step 2, by setting the amplitude threshold, convert the waveform into a square wave form, which is convenient for calculating the period with the falling edge as the characteristic point in subsequent steps.

[0078] 3.2 The threshold is usually set near zero. To ensure sufficient adaptability to wave valleys of different amplitudes, the filtered signal can be amplified by an appropriate multiple in implementation, and then a small negative threshold is selected below zero. When the pulse signal is higher than this threshold, it is marked as high level, otherwise it is low level.

[0079] 3.3 Denote the high level and low level as different Boolean or integer identifiers respectively; in the case of continuous acquisition, each newly acquired data signal is compared with the threshold in real-time, and the result is updated to a fixed-length array or buffer for the next cycle detection.

[0080] Step 4: Period detection and calculation of the main wave interval of the pulse

[0081] 4.1 Determine the end of each pulse cycle by detecting the falling edge of the square wave signal. When the difference between the current time flag of the square wave and the previous time flag is negative, it can be considered that the falling edge has occurred.

[0082] 4.2 While detecting the falling edge, it is necessary to confirm whether the duration of the high-level segment conforms to the normal pulse period. If the number of sampled points of the high level is too small, it may be instantaneous interference or noise, and this segment of waveform needs to be discarded; if it reaches a certain threshold, this segment of waveform is regarded as a valid pulse period.

[0083] 4.3 Calculate the detection time difference between two adjacent valid pulse periods, and the pulse main wave interval (which can also be denoted as the PP interval) can be obtained. To improve real-time performance, the information such as the position and width of the most recent peak is stored in a circular sliding window or a queue, and is updated each time a valid falling edge is confirmed.

[0084] Step 5: Data Output and Real-time Application

[0085] 5.1 The pulse main wave interval obtained in Step 4 can be further used to calculate various physiological indexes such as heart rate variability in real time, or can be directly used as a reference for detecting the heart rate or rhythm state.

[0086] 5.2 Output the timing result of the pulse main wave interval through the display interface of the mobile terminal or wearable device, or transmit it to the host computer through the wireless communication module to provide users with instant cardiovascular health monitoring data.

[0087] 5.3 In specific applications, the above steps can be integrated into a microprocessor, a smartphone APP, a smart watch or other wearable device software according to requirements to realize long-term monitoring of pulse signals and early warning of abnormal events.

[0088] III. Feasibility and Improvement Description

[0089] This solution is not only applicable to resting or light exercise scenarios, but can also cope with moderate-intensity daily activities after appropriately adjusting the sensor wearing method and filtering parameters.

[0090] The specific parameters of the filtering algorithm and threshold detection can be appropriately fine-tuned according to the characteristics of the user population (such as the elderly, children, athletes, etc.) to ensure robustness in different physiological conditions and measurement environments.

[0091] To improve the scalability of this solution, other auxiliary decision-making logics (such as pulse waveform similarity discrimination) can be added in addition to baseline drift elimination and falling edge detection to further reduce the influence of motion artifacts or abnormal light interference.

[0092] If there is higher hardware performance support, more filtering levels or machine learning algorithms can be combined to strengthen the capture and analysis of subtle pulse characteristics.

[0093] Implementation Method III. Combination Figure 1-8To describe this embodiment, the core principle of this embodiment lies in the innovative development of a real-time pulse main wave interval sequence calculation technology for mobile pulse signals, which is based on a single-point real-time pulse main wave interval sequence extraction algorithm. First, a full-wrap finger clip is used to continuously collect the pulse signal in the region of interest, and then an infrared signal filtering method based on a second-order IIR (Infinite Impulse Response) filter is adopted, and a notch filtering and baseline drift collaborative elimination algorithm based on integer coefficient optimization is used to process the original pulse signal in real time. For the processed pulse signal, a real-time pulse main wave interval sequence detection algorithm is applied, and a threshold is set to quantize the pulse signal into a square wave.

[0094] Example 1:

[0095] As Figure 1 shown, a single-point real-time pulse main wave interval sequence extraction method includes the following steps:

[0096] S01: Collect the human pulse signal through a full-wrap photoplethysmography technology acquisition system, greatly reducing external light interference;

[0097] S02: Set an infrared signal filtering method based on a second-order infinite impulse response filter to preprocess the signal in real time, and use a notch filtering and baseline drift collaborative elimination algorithm based on integer coefficient optimization to filter the preprocessed signal in real time;

[0098] S03: This embodiment proposes a single-point real-time pulse main wave interval sequence extraction algorithm, quantize the clean wave after real-time filtering into a square wave through this algorithm, calculate the arrival of the falling edge, and perform the calculation of real-time pulse main wave interval sequence extraction.

[0099] In a preferred embodiment, the data processing method of S02 is as Figure 2 shown. This embodiment is based on the original signal data processing flow chart. The following steps are included in step S02:

[0100] Parse the data of the continuously collected original pulse signal data stream to extract the infrared signal data. Formula:

[0101] data_ir = (B2 × 216 + B1 × 28 + B0) × scaling_factor

[0102] In the formula, B2, B1, and B0 are the high-order to low-order bytes, which need to be combined according to the weight, and scaling_factor is the sensor sensitivity.

[0103] Normalize the data and map the original value to the standard range:

[0104]

[0105] μ baseline is the mean of the baseline period data, and σ baseline is the standard deviation of the baseline period data.

[0106] In this preferred embodiment, an infrared signal filtering method based on an infinite impulse response filter is used to suppress and remove the tiny glitch signals on its original pulse signal. By designing the forward coefficients and feedback coefficients of the filter and adopting a recursive algorithm to perform step-by-step processing of the signal.

[0107] The mathematical expression of the filter is:

[0108] y[n] = b0·x[n] + b1·x[n - 1] + b2·x[n - 2]

[0109] - a1·y[n - 1] - a2·y[n - 2]

[0110] x[n] is the current input signal, y[n] is the current output signal, b0, b1, b2 are the forward coefficients of the filter, and a1, a2 are the feedback coefficients of the filter.

[0111] Forward coefficients: b = {0.0036, 0.0072, 0.0036}

[0112] Feedback coefficients: a = {1.0, -1.8227, 0.8372}

[0113] First, perform initialization, pre-store x[n - 1], x[n - 2], y[n - 1], y[n - 2], perform recursive calculation, update the output y[n] according to the difference equation, and then update the historical values x[n - 2] ← x[n - 1], x[n - 1] ← x[n], and similarly update y[n].

[0114] In a preferred embodiment, the preprocessing method in S02 further includes: a notch filtering and baseline drift co-elimination algorithm based on integer coefficient optimization.

[0115] This embodiment proposes a cascaded integer coefficient filtering architecture, which includes the following key modules:

[0116] Module 1: Design of integer coefficient notch filter:

[0117] Frequency response characteristic modeling: According to the target noise frequency band, establish the transfer function:

[0118]

[0119] where θ = 2πf / fs, f is the target suppression frequency, and r is the pole radius.

[0120] Integer approximation optimization:: 1. Coefficient scaling: Multiply both the numerator and denominator of the transfer function by 2 N , to convert the key coefficients into integers. 2. Shift substitution: Use bit shift operations to implement 2 N division. 3. Stability verification: The pole positions need to satisfy |r| < 1 to ensure the stability of the quantized system.

[0121] Module 2: Baseline drift dynamic elimination:

[0122] Sliding window difference: Implemented using a circular buffer of length L (in the embodiment L = 641).

[0123] PPG_IR = {PPG1, PPG2,..., PPG n ,..., PPG L}

[0124] PPG n is the nth sampling point within the window, PPG1 is the earliest sampling point, and the newly collected sampling point is placed at the highest bit in the window buffer, denoted as PPG L . Each time the data is updated, the sampling points shift one bit from the high bit to the low bit, the earliest collected PPG1 is covered by the high bit, and the new data is placed at PPG L . Calculate the difference:

[0125] Δ trend = PPG_IR

[641] - PPG_IR[1]

[0126] Trend term extraction: Separate the low-frequency baseline drift through the difference operation of the long-period components within the window:

[0127]

[0128] Drift compensation: Output signal correction:

[0129] y corrected [n] = y[n] - Trend[n]

[0130] Collaborative filtering architecture:

[0131] Cascaded execution order: First perform rough elimination of baseline drift, and then implement notch filtering to avoid amplification of high-frequency noise during trend term extraction.

[0132]

[0133] Feedback compensation mechanism: Realize closed-loop error correction by caching historical output values in the Y_IR array.

[0134] y[n] = NotchFilter(x[n] - β·Y IR [n - 1])

[0135] where β is the feedback gain to suppress the residual error, and Y IR [n - 1]) is the data point immediately preceding the current data point of the Y_IR array.

[0136] Cascade baseline cancellation and notch filtering:

[0137] H total (z) = H baseline (z) · H notch (z)

[0138] H baseline (z) is the moving window difference to cancel the baseline drift.

[0139] In a preferred embodiment, a method for accurately calculating the real-time pulse main wave interval sequence of the standard pulse wave is proposed in S03. The overall flowchart is as Figure 3 shown, and the specific method is as follows:

[0140] Quantize the filtered standard pulse wave into a square wave, equivalent to high and low levels. Obtain the filtered pulse signal. The filtered pulse signal can be regarded as an approximate periodic signal. According to the waveform characteristics of the dynamic pulse signal, an amplitude threshold method is proposed to extract the pulse main wave interval sequence signal from the pulse signal in real time. The input signal y[n] is centered near zero, and the output square wave sequence pulse(n) ∈ {0, 1}.

[0141] As Figure 6 shown, Figure 6 is a schematic diagram of the real-time pulse main wave interval sequence extraction algorithm;

[0142] y scaled (n) = y(n) × 1000

[0143]

[0144] pulse(n - 1) ← pulse(n)

[0145] In this embodiment, the signal value obtained in real time is scaled up by 1000 times. Set a threshold Th1 to perform binarization on the data, and dynamically quantize the pulse signal into a square wave signal of 0 or 1. When the threshold Th1 is selected as -1, the resulting data error can be almost negligible. Subsequently, perform a moving window update on the data pulse(n) for edge detection, record the square wave signal, shift the previous data forward, and put the real-time data into pulse(n), processing one by one.

[0146] In a preferred embodiment, the algorithm proposed in S03 further includes feature extraction: calculating the widths of the high and low levels of the pulse signal quantized into a square wave sequence, calculating the main wave interval based on this as a reference, inputting the square wave sequence pulse(n), and outputting the high level width H, the low level width L, and the peak position MaxLoc:

[0147] Judge each pulse data point collected and processed in real time. The initial values are H(0) = 0 and L(0) = 0. Detect and judge the value of the current data point pulse(n) in real time. When pulse(n) = 1, the width H of the high level is incremented by 1. When pulse(n) = 0, the width L of the low level is incremented by 1. After that, the above process is performed for each data point collected.

[0148]

[0149] Peak position tracking: The initial conditions are peak Max = 0 and peak position MaxLoc = 0. Record and judge the current single data point y scaled (n) with Max. If it is greater than Max, update the peak and the peak position; otherwise, keep it unchanged.

[0150]

[0151] As Figure 7 shown, this embodiment uses the rising and falling edges of the high and low levels for calculation, and adopts a sliding window method to calculate and detect the arrival of the falling edge to calculate the real-time pulse main wave interval sequence:

[0152] When the difference between the current pulse signal and the previous pulse signal is -1, it can be known that the falling edge of the square wave arrives. In addition, it is necessary to judge the high level width. The sampling frequency is 500Hz. If the high level width is less than 100, this section of the signal is an interference signal section that needs to be removed.

[0153]

[0154] When the arrival of the falling edge is detected, calculate the pulse main wave interval, that is, the sum of the high level width in the latter part of the current cycle, the peak offset of the previous cycle, and the low level width is the pulse main wave interval width. In the formula, H1 and H2 are arrays of length 2.

[0155]

[0156] Then, extract the main wave interval sequence according to the pulse main wave interval width:

[0157]

[0158] Finally, reset the sliding window to initialize the parameters for the next cycle calculation:

[0159] H←0,L←0

[0160] Max ← 0, MaxLoc ← 0

[0161] H1[1] ← H1[2], H2[1] ← H2[2]

[0162] In another embodiment, a computer storage medium stores a computer program, and when the computer program is executed, the above-mentioned method for extracting the pulse main wave interval sequence in real time at a single point is implemented.

[0163] In yet another embodiment, this embodiment also discloses a system for extracting the pulse main wave interval sequence in real time at a single point. As Figure 4 shown in the structural schematic diagram of the real-time pulse main wave interval sequence extraction system for mobile pulse signal processing and algorithms, it includes:

[0164] A pulse signal acquisition module 10, which uses an all-inclusive finger clip to isolate light source interference, acquires two-dimensional pulse signals, and preprocesses the acquired pulse signals.

[0165] A pulse signal processing module 20, which is specially designed for filtering and algorithm processing of pulse signals, can effectively reduce the noise interference of pulse signals. Through this key step, the module can effectively remove the tiny burrs on the pulse signals and the baseline drift of the pulse signals, ensuring the accuracy and reliability of the subsequent data processing steps.

[0166] A pulse main wave interval sequence extraction module 30, which quantifies the processed pulse signals into square waves, calculates the high and low level widths of the square waves, determines the pulse width of one cycle using the arrival of the falling edge of the square wave, and uses it as the PP interval, thereby realizing the extraction of the pulse main wave interval sequence in real time at a single point.

[0167] Specifically, taking a preferred embodiment as an example below, the system for extracting the pulse main wave interval sequence in real time at a single point and its working process Figure 5 is the overall flowchart of the system for extracting the pulse main wave interval sequence in real time at a single point, and is specifically described as follows:

[0168] mainly includes the following steps:

[0169] Step 1: Through an all-inclusive pulse signal acquisition system, collect the region of interest at the fingertip. However, it should be noted that the target of the all-inclusive acquisition sensor is the fingertip, and the sensor cannot be worn too deeply to prevent light from entering the sensor and causing interference. The specific steps are as follows:

[0170] (1) Adjust the wearing depth of the all-inclusive sensor to prevent light from entering and causing interference;

[0171] (2) Connect the sensor to the acquisition device and connect it to the upper acquisition software or system through the type-c serial port protocol;

[0172] (3) For the convenience of users' operation, parameters such as sampling frequency have been set in advance. It is only necessary to control whether to collect pulse signals, the mode of collecting pulse signals, etc. in the acquisition software.

[0173] Step 2: Set the corresponding filters and algorithms to process the collected pulse signals in real time. The specific steps are as follows:

[0174] Step 2.1: Parse the data of the continuously collected original pulse signal data stream to extract the infrared signal data. Formula:

[0175] data_ir = (B2 × 2^16 + B1 × 2^8 + B0) × scaling_factor

[0176] In the formula, B2, B1, and B0 are bytes from high to low and need to be combined according to weights. scaling_factor is the sensor sensitivity.

[0177] Normalize the data and map the original value to the standard range:

[0178]

[0179] μ baseline is the mean value of the data in the baseline period, and σ baseline is the standard deviation of the data in the baseline period.

[0180] Step 2.2: Use an infrared signal filtering method based on an infinite impulse response filter to suppress and remove the tiny glitch signals on its original pulse signal. By designing the forward coefficients and feedback coefficients of the filter and using a recursive algorithm to process the signals step by step.

[0181] The mathematical expression of the filter is:

[0182] y[n] = b0 · x[n] + b1 · x[n - 1] + b2 · x[n - 2]

[0183] - a1 · y[n - 1] - a2 · y[n - 2]

[0184] x[n] is the current input signal, y[n] is the current output signal, b0, b1, b2 are the forward coefficients of the filter, and a1, a2 are the feedback coefficients of the filter.

[0185] Forward coefficients: b = {0.0036, 0.0072, 0.0036}

[0186] Feedback coefficients: a = {1.0, -1.8227, 0.8372}

[0187] First, perform initialization, pre-store x[n - 1], x[n - 2], y[n - 1], y[n - 2], perform recursive calculation, update the output y[n] according to the difference equation, and then update the historical values x[n - 2] ← x[n - 1], x[n - 1] ← x[n], and update y[n] in the same way.

[0188] Step 2.3: Notch filtering optimized based on integer coefficients and collaborative elimination algorithm for baseline drift.

[0189] Propose a cascaded integer coefficient filtering architecture, which includes the following key points:

[0190] Key point 1: Integer coefficient notch filter.

[0191] Frequency response characteristic modeling: According to the target noise frequency band, establish the transfer function:

[0192]

[0193] where θ = 2πf / fs, f is the target suppression frequency, and r is the pole radius.

[0194] Integer approximation optimization: 1. Coefficient scaling: Multiply both the numerator and denominator of the transfer function by 2 N , so that the key coefficients are converted into integers. 2. Shift substitution: Use bit shift operations to implement 2 N division. 3. Stability verification: The pole position needs to satisfy |r| < 1 to ensure the stability of the system after quantization.

[0195] Key point 2: Dynamic elimination of baseline drift.

[0196] Sliding window difference: Implemented using a circular buffer with a length of L (L = 641 in the embodiment).

[0197] PPG_IR = {PPG1, PPG2,..., PPG n ,..., PPG L}

[0198] PPG n is the nth sampling point in the window, PPG1 is the earliest sampling point, and the latest collected sampling point is placed at the highest bit in the window buffer, denoted as PPG L , and each time the data is updated, the sampling points move one bit from the high bit to the low bit. The earliest collected PPG1 is covered by the high bit, and the new data is placed at PPG L . Calculate the difference:

[0199] Δ trend = PPG_IR

[641] - PPG_IR[1]

[0200] Trend term extraction: Separate the low-frequency baseline drift through the difference operation of the long-period components within the window:

[0201]

[0202] Drift compensation: Output signal correction:

[0203] y corrected [n] = y[n] - Trend[n]

[0204] Key point 3: Collaborative filtering architecture.

[0205] Cascaded execution order: First, perform rough elimination of baseline drift, and then implement notch filtering to avoid amplification of high-frequency noise during trend term extraction.

[0206]

[0207] Feedback compensation mechanism: Cache historical output values through the Y_IR array to achieve closed-loop error correction.

[0208] y[n] = NotchFilter(x[n] - β·Y IR [n - 1])

[0209] where β is the feedback gain to suppress residual error, and Y IR [n - 1]) is the data point before the current data point of the Y_IR array.

[0210] Cascade baseline elimination and notch filtering:

[0211] H total (z) = H baseline (z)·H notch (z)

[0212] H baseline (z) is the sliding window difference to eliminate baseline drift.

[0213] Step 3: A method for accurately calculating the real-time pulse main wave interval sequence of a standard pulse wave. Step 3 is divided into the following sub-steps:

[0214] Step 3.1:

[0215] Quantize the filtered standard pulse wave into a square wave, equivalent to high and low levels. Obtain the filtered pulse signal. The filtered pulse signal can be regarded as an approximate periodic signal. According to the waveform characteristics of the dynamic pulse signal, an amplitude threshold method is proposed to extract the pulse main wave interval sequence signal from the pulse signal in real time. The input signal y[n] is centered near zero, and the output square wave sequence pulse(n) ∈ {0, 1}.

[0216] As Figure 6 shown, Figure 6Schematic diagram of the real-time pulse main wave interval sequence extraction algorithm;

[0217] y scaled (n) = y(n) × 1000

[0218]

[0219] pulse(n - 1) ← pulse(n)

[0220] The numerically acquired real-time signal is enlarged by a factor of 1000. Set a threshold Th1 and perform binarization processing on the data to dynamically quantize the pulse signal into a square wave signal of 0 or 1. When the threshold Th1 is selected as -1, the resulting data error can be almost ignored. Subsequently, perform a sliding window update on the data pulse(n) for edge detection, record the square wave signal, shift the previous data forward, put the real-time data into pulse(n), and process them one by one.

[0221] Step 3.2:

[0222] Feature extraction: Calculate the widths of the high and low levels of the pulse signal quantized into a square wave sequence, and calculate the main wave interval based on this. Input the square wave sequence pulse(n), and output the high-level width H, the low-level width L, and the peak position MaxLoc:

[0223] Judge each pulse data point collected and processed in real time. The initial values are H(0) = 0 and L(0) = 0. Real-time detect and judge the value of the current data point pulse(n). When pulse(n) = 1, increment the high-level width H by 1. When pulse(n) = 0, increment the low-level width L by 1. After that, perform the above process for each collected data point.

[0224]

[0225] Peak position tracking: The initial conditions are peak Max = 0 and peak position MaxLoc = 0. Record and judge the current single data point y scaled (n) against Max. If it is greater than Max, update the peak and the peak position; otherwise, keep them unchanged.

[0226]

[0227] Step 3.3:

[0228] Adopt a sliding window method to calculate and detect the arrival of the falling edge to calculate the real-time pulse main wave interval sequence:

[0229] When the result of subtracting the current pulse signal from the previous pulse signal is -1, it can be known that the falling edge of the square wave arrives. In addition, it is necessary to judge the high-level width. The sampling frequency is 500Hz. If the high-level width is less than 100, this section of the signal is an interference section signal that needs to be removed.

[0230]

[0231] When the arrival of the falling edge is detected, the calculation of the pulse main wave interval is performed, that is, the sum of the high-level width in the latter segment of the current cycle, the peak offset of the previous cycle, and the low-level width is the pulse main wave interval width. In the formula, H1 and H2 are arrays with a length of 2.

[0232]

[0233] Then, the main wave interval sequence is extracted according to the pulse main wave interval width:

[0234]

[0235] Finally, the sliding window is reset to initialize the parameters for the next cycle calculation:

[0236] H←0,L←0

[0237] Max←0,MaxLoc←0

[0238] H1[1]←H1[2],H2[1]←H2[2]

[0239] The extraction result of the main wave interval sequence is as Figure 8 shown.

[0240] The core processing is shown in Table 1. This process realizes the fully automatic real-time analysis from the original signal to the heart rate parameters.

[0241] Table 1

[0242]

[0243] Furthermore, the signal acquisition lower computer consists of a two-dimensional pulse data detection system and a microprocessor. The two-dimensional pulse data detection system is used to acquire the pulse signal at the fingertip part of the human body. The microprocessor filters the signal. The microprocessor can be a single-chip microcomputer, DSP, etc., but is not limited to these processors.

[0244] Furthermore, the communication method of the serial communication module is through data signal lines, ground wires, control lines, etc., but is not limited to these communication methods.

[0245] Furthermore, the upper computer for signal processing can be a microprocessor, a wearable device, a PC, or a smart phone. The completed single-point real-time pulse main wave interval sequence extraction system can be embedded in a microprocessor, a wearable device, a PC, or a smart phone to realize the extraction of the pulse main wave interval sequence. The microprocessor can be a processor of the type such as a single-chip microcomputer or a DSP, but is not limited to these processors.

[0246] The above further describes the technical solutions provided by the present invention through several specific embodiments to highlight the advantages and beneficial effects of the technical solutions provided by the present invention. However, the above-mentioned several specific embodiments are not used as a limitation to the present invention. Any reasonable modifications and improvements, combinations of embodiments, equivalent replacements, etc. based on the spirit and principle of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for extracting a single-point real-time pulse main wave interval sequence, characterized in that, Including the following steps: The step of obtaining a two-dimensional pulse signal to be processed by using an all-inclusive photoplethysmogram acquisition system; The step of performing multi-byte data parsing on the two-dimensional pulse signal to obtain a pulse signal data stream for subsequent operations; The step of performing second-order IIR filtering on the pulse signal data stream and using a cascaded integer coefficient notch filter and baseline drift collaborative elimination algorithm for real-time noise reduction and baseline correction, and outputting a standard pulse wave; The step of setting an amplitude threshold for the standard pulse wave and converting the waveform into a square wave to obtain a high and low level signal; The step of detecting the falling edge of the square wave signal to determine the end of each pulse cycle and filtering out noise pulses with insufficient high-level width; The step of calculating the time difference between adjacent effective pulse cycles and real-time outputting a pulse main wave interval sequence.

2. The method for extracting the single-point real-time pulse main wave interval sequence according to claim 1, wherein, The all-inclusive photoplethysmogram acquisition system adopts a finger clip structure, which can form a closed space between the sensor and the finger to reduce external light interference.

3. A method for extracting the main wave interval sequence of a single-point real-time pulse according to claim 1, characterized in that, The second-order IIR filtering preferentially suppresses high-frequency noise and burr signals, and provides a smoother input signal for subsequent integer coefficient notch filtering.

4. A method for extracting the main wave interval sequence of a single-point real-time pulse according to claim 1, characterized in that, The baseline drift collaborative elimination algorithm dynamically corrects the slow-changing trend of the pulse waveform in a sliding window manner to ensure that the overall mean value of the pulse waveform is within a set interval.

5. A method for extracting the main wave interval sequence of a single-point real-time pulse, characterized in that, The quantization step magnifies the standard pulse wave by a multiple and selects a threshold near zero to more accurately distinguish the high-level section and the low-level section.

6. A method for extracting a single-point real-time pulse main wave interval sequence according to claim 1, characterized in that, The cycle detection step determines an effective pulse cycle by judging the high-level width, and only judges it as an effective pulse cycle when the width exceeds a preset threshold to eliminate false pulses caused by motion artifacts or sudden noises.

7. A single-point real-time device for extracting the pulse main wave interval sequence, characterized in that, Including the following modules: A module for obtaining a two-dimensional pulse signal to be processed by using an all-inclusive photoplethysmogram acquisition system; A module for performing multi-byte data parsing on the two-dimensional pulse signal to obtain a pulse signal data stream for subsequent operations; A module for performing second-order IIR filtering on the pulse signal data stream and using a cascaded integer coefficient notch filter and baseline drift collaborative elimination algorithm for real-time noise reduction and baseline correction, and outputting a standard pulse wave; A module for setting an amplitude threshold for the standard pulse wave and converting the waveform into a square wave to obtain a high and low level signal; A module for detecting the falling edge of the square wave signal to determine the end of each pulse cycle and filtering out noise pulses with insufficient high-level width; A module for calculating the time difference between adjacent effective pulse cycles and real-time outputting a pulse main wave interval sequence.

8. A computer storage medium for storing a computing program, characterized in that, When the computer program is read by a computer, the computer executes the method described in claim 1.

9. A computer, comprising a processor and a storage medium, characterized in that, When the processor reads the computer program stored in the storage medium, the computer executes the method described in claim 1.

10. A computer program product, as a computer program, characterized in that, When the computer program is executed, the method described in claim 1 is implemented.

Images