A Millimeter-Wave Radar Heart Rate Measurement Method Based on Adaptive Notch Filtering

Abstract

This invention discloses a millimeter-wave radar method for measuring human heart rate based on adaptive notch filtering. The invention achieves stable temporal phase waveform extraction by compensating for phase jumps across distance cells caused by random human body movement and performing phase unwrapping. An adaptive notch filter is designed by extracting the respiratory fundamental frequency and bandwidth characteristics, and combined with adaptive differential processing based on local signal-to-noise ratio to obtain a filtered and differentially enhanced phase waveform. The harmonic product spectrum of the enhanced waveform is obtained, and the validity of the measurement value is judged based on confidence level. The current heart rate is then dynamically updated using the filter.

Description

Technical Field

[0001] This invention belongs to the field of millimeter-wave radar technology, specifically, it relates to a millimeter-wave radar method for measuring human heart rate based on adaptive notch filtering. Background Technology

[0002] As a non-contact method for monitoring vital signs, millimeter-wave radar has been widely used in the field of health management. Millimeter-wave radar can capture raw echoes containing micron-level chest wall displacement information. By performing pulse compression on the raw echoes, extracting complex data of the distance cell where the target human body is located, and performing phase unwrapping, the chest wall displacement curve can be obtained, thereby achieving accurate estimation of heart rate parameters.

[0003] The current technical challenges of millimeter-wave radar for human heart rate measurement are mainly reflected in three aspects: First, random swaying of the human body may cause changes in the distance between the millimeter-wave radar and the human chest cavity across resolution cells, resulting in distortion of the timing phase waveform; Second, heartbeat and respiratory waveforms are mixed in the time and frequency domains, and conventional filtering algorithms are difficult to effectively separate heartbeat and respiratory components; Third, the amplitude of human respiration is significantly greater than that of heartbeat, and weak heartbeat signals are easily masked by respiratory harmonics.

[0004] The aforementioned problems make it difficult for traditional methods to achieve stable and accurate heart rate measurement, and there is an urgent need to design and implement a more adaptable heart rate measurement algorithm. Summary of the Invention

[0005] In view of this, the present invention provides a millimeter-wave radar method for measuring human heart rate based on adaptive notch filtering, which can achieve stable, accurate and reliable measurement of the heart rate of a stationary person.

[0006] The technical solution for implementing the present invention is as follows: A millimeter-wave radar method for measuring human heart rate based on adaptive notch filtering includes: Step 1: For a one-dimensional range image of millimeter-wave radar, compensate for the phase jump across range cells caused by random human body movement, and extract the time-series phase waveform after phase unwrapping. Step 2: Based on the phase waveform spectrum, extract the respiratory fundamental frequency and bandwidth characteristics, design an adaptive notch filter, and complete the filtering process of the phase waveform; Step 3: Perform first-order and second-order differential processing on the filtered phase waveform. Select the differential order according to the ratio of local signal-to-noise ratio to obtain the differentially enhanced phase waveform. Step 4: Obtain the harmonic product spectrum of the differential enhancement waveform, determine the validity of the current heart rate measurement based on the confidence level, and combine it with... The filter enables dynamic updates to the current heart rate.

[0007] In step one, the phase term that needs to be compensated before phase unwrapping is:

[0008] in, For distance units, This represents the number of sampling points.

[0009] In step two, after obtaining the 3dB bandwidth based on the waveform spectrum, the formula for obtaining the frequency doubling order is:

[0010] in, For the frequency multiplication factor, It is the fundamental frequency of respiration. The duration of a single breath.

[0011] In step three, the formula for calculating the order selection value is:

[0012] in, Let be the local mean of the maximum peak in the spectrum of order i. Let be the local variance of the maximum peak in the spectrum of order i.

[0013] In step four, the confidence level is calculated using the following formula:

[0014] in, The amplitude of the heart rate measurement obtained from the differential signal in the harmonic product spectrum. This represents the amplitude of the maximum peak in the harmonic product spectrum.

[0015] Beneficial effects 1. Significantly improves the ability to separate heartbeat and respiratory signals under complex interference; 2. Effectively overcomes phase distortion caused by random body movements, significantly enhancing the environmental robustness of heart rate measurement; 3. A screening mechanism based on confidence assessment is proposed to ensure the high stability and high reliability of the measurement results. Attached Figure Description

[0016] Figure 1 A flowchart illustrating the implementation of this invention; Figure 2 Comparison diagram of phase compensation; Figure 3 Schematic diagram of adaptive parameters for notch filter; Figure 4 A schematic diagram of obtaining the confidence level of the measured value using the harmonic product spectrum. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings.

[0018] This invention provides a millimeter-wave radar method for measuring human heart rate based on adaptive notch filtering, comprising the following steps: Step 1: For a one-dimensional range image of millimeter-wave radar, compensate for the phase jump across range cells caused by random human body movement, and extract the time-series phase waveform after phase unwrapping.

[0019] Millimeter-wave radar typically uses the FMCW signal system, and the transmitted and received signals can be represented as follows: (1) (2) in, For carrier frequency, For scan cycle, For bandwidth, For distance, For delay and To adjust the frequency.

[0020] The intermediate frequency signal after conjugate multiplication is: (3) The signal after discrete sampling is: (4) in, The number of sampling points. The position of the initial phase.

[0021] After Fourier transform: (5) in, For distance units.

[0022] From formula (5), we can see that the Fourier transform signal contains five phase terms, among which the fifth phase term is directly related to the range cell where the target is located. The fifth phase term is first processed on the complex data of the peak points. Phase compensation is performed, followed by phase unwrapping of the compensated complex data, which corrects the phase jump caused by the peak point spanning multiple cells and restores the phase waveform. The compensation effect is as follows: Figure 2 As shown.

[0023] Step 2: Based on the phase waveform spectrum, extract the respiratory fundamental frequency and bandwidth characteristics, design an adaptive notch filter, and complete the filtering process of the phase waveform.

[0024] A Fourier transform is performed on the unwrapped phase waveform, and the peak value of the respiratory fundamental frequency is detected in the 0.2-0.4Hz frequency range. The 3dB bandwidth (i.e., the amplitude drops to its maximum value) at this frequency is obtained based on the spectral amplitude characteristics. The frequency range of time, such as Figure 3 (As shown). Based on this respiratory fundamental frequency and bandwidth, a set of stopband center frequencies is generated. The multi-stage notch filter is finally set to the stopband bandwidth of the notch filter.

[0025] (6) in, For the frequency multiplication factor, It is the fundamental frequency of respiration. The duration of a single breath.

[0026] Step 3: Perform first-order and second-order differential processing on the filtered phase waveform. Select the differential order based on the ratio of the local signal-to-noise ratio to obtain the differentially enhanced phase waveform.

[0027] Ideal thoracic cavity displacement waveform Based solely on respiratory waveforms and heart rate waveform Composition, namely: (7) in, This refers to the range of breathing. This refers to the amplitude of the heartbeat. The fundamental frequency of the heartbeat, The phase of the respiratory waveform. This represents the phase of the heartbeat waveform.

[0028] The first-order and second-order differential time series of the thoracic displacement waveform are as follows: (8) (9) It can be seen that the differentiation operation has little effect on the frequency of each signal component, but a significant effect on the amplitude. Specifically, after the first derivative, the amplitude change of the heartbeat component is significantly greater than that of the respiratory component, reaching its maximum. This difference is even more pronounced in the second-order differential, where the amplitude variation of the heartbeat component is far greater than that of the respiratory component, reaching its [value missing]. The difference is significant. Clearly, differential processing can significantly enhance the heartbeat signal, especially its higher harmonic components, making it a very effective method for enhancing weak heartbeat signals.

[0029] However, in the original waveform of the measured data, the second harmonic component of heart rate (2-4Hz band) itself has relatively weak energy, and its intensity is often comparable to that of noise components. While second-order differential processing enhances the heartbeat signal, it may also amplify other interference components. Therefore, it is necessary to select appropriate differential signals based on the spectral characteristics of the first and second-order signals respectively.

[0030] This invention proposes a selection criterion based on the ratio of the local signal-to-noise ratio (SNR) of the differential signals: when the difference between the peak SNR of the first-order differential signal and the peak SNR of the second-order differential signal exceeds an empirical threshold, the first-order differential signal is deemed to have higher reliability, and the first-order differential result will be used; otherwise, the second-order differential result will be used. The final heart rate measurement is determined by the maximum spectral peak value within the corresponding frequency band (0.8-2 Hz or 2-4 Hz) of the selected differential signal.

[0031] The formula for calculating the local signal-to-noise ratio difference is: (10) in, Let be the local mean of the maximum peak in the spectrum of order i. Let be the local variance of the maximum peak in the spectrum of order i.

[0032] Step 4: Obtain the harmonic product spectrum of the differential enhancement waveform, determine the validity of the current heart rate measurement based on the confidence level, and combine it with... The filter enables dynamic updates to the current heart rate.

[0033] Multiplying the spectrum of the differential signal by the spectrum of its upsampled signal (which is twice as high) yields the harmonic product spectrum, such as... Figure 4 As shown.

[0034] In the harmonic product spectrum, the heart rate confidence level is obtained by calculating the ratio of the heart rate measurement obtained from the differential signal to the maximum peak amplitude in the harmonic product spectrum. When the confidence level exceeds a preset threshold, the current heart rate measurement is considered reliable. The confidence level calculation formula is as follows: (11) in, The amplitude of the heart rate measurement obtained from the differential signal in the harmonic product spectrum. This represents the amplitude of the maximum peak in the harmonic product spectrum.

[0035] When the measured value is reliable, use The filter performs fusion filtering on the measured value and the historical value, then outputs the filtered value and uses the filtered value to update the historical value; when the measured value is unreliable, the historical value is output.

[0036] Ten subjects with normal cardiac function (6 men and 4 women) were recruited for the experiment. Subjects remained seated at a distance of approximately 0.7–1.0 meters from the radar. Heart rate data acquired using a Xiaomi Mi Band was used as the reference value. The total data collection time was 80 minutes, resulting in 485 valid measurements.

[0037] To evaluate the experimental performance of the proposed method, we use mean absolute error (MAE) and root mean square error (RMSE) as evaluation metrics, defined as follows:

[0038]

[0039] Where W represents the number of time windows within the observation period. and The first The true value and estimated value of each time window.

[0040]

[0041] The measurement results are shown above. The absolute error and root mean square error of the heart rate estimates of the 10 subjects were within 3 bpm. This method has a good effect on heart rate measurement under static sitting conditions.

[0042] In summary, the above are merely preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for measuring human heart rate using millimeter-wave radar based on adaptive notch filtering, characterized in that, include: Step 1: For a one-dimensional range image of millimeter-wave radar, compensate for the phase jump across range cells caused by random human body movement, and extract the time-series phase waveform after phase unwrapping. Step 2: Based on the phase waveform spectrum, extract the respiratory fundamental frequency and bandwidth characteristics, design an adaptive notch filter, and complete the filtering process of the phase waveform; Step 3: Perform first-order and second-order differential processing on the filtered phase waveform. Select the differential order according to the ratio of local signal-to-noise ratio to obtain the differentially enhanced phase waveform. Step 4: Obtain the harmonic product spectrum of the differential enhancement waveform, determine the validity of the current heart rate measurement based on the confidence level, and combine it with... The filter enables dynamic updates to the current heart rate.

2. The method as described in claim 1, characterized in that, In step one, the discrete sampled signal after Fourier transform is: ； in, For distance units, The number of sampling points. The initial frequency, For bandwidth, The slope; The fifth phase term is first applied to the complex data at the peak point. Phase compensation.

3. The method as described in claim 1, characterized in that, In step two, a Fourier transform is performed on the unwrapped phase waveform, and the 3dB bandwidth corresponding to the breathing fundamental frequency is obtained according to the spectral amplitude characteristics. Based on the breathing fundamental frequency and the 3dB bandwidth, a set of multi-stage notch filters with a stopband center frequency that is an integer multiple of the breathing fundamental frequency is generated, and the 3dB bandwidth is set as the stopband bandwidth of the notch filter to filter the phase waveform.

4. The method as described in claim 1, characterized in that, In step three, the formula for calculating the order selection value is: ； in, Let be the local mean of the maximum peak in the spectrum of order i. Let be the local variance of the maximum peak in the spectrum of order i.

5. The method as described in claim 1, characterized in that, In step four, the confidence level is calculated using the following formula: ； in, The amplitude of the heart rate measurement obtained from the differential signal in the harmonic product spectrum. This represents the amplitude of the maximum peak in the harmonic product spectrum.

6. The method as described in claim 1, characterized in that, In step four, the harmonic product spectrum of the differential enhancement waveform is obtained, the validity of the current heart rate measurement value is determined based on the confidence level, and the current heart rate is dynamically updated in conjunction with the filter.

Images