2.1

PVDF piezoelectric sensor sleep respiratory detection

Some crystalline materials exhibit piezoelectric effects[7] under external forces, accumulating charges on their surfaces or within, thereby generating voltage. By encapsulating such materials into PVDF piezoelectric film sensors, highly sensitive sensors can detect subtle pressure changes.

Placing the PVDF piezoelectric film sensor between a mattress and the human body, respiratory movements cause slight deformations in the film[8], resulting in changes in electric charge. The mathematical relationship between the output voltage V and the applied force F is expressed as Equation (1) (where d is the piezoelectric constant, and C is the capacitance connected to the piezoelectric element). The small amount of charge output by the piezoelectric sensor under stress is further amplified by auxiliary signal conditioning circuits to capture and convert respiratory signals. Finally, signals are filtered and separated to extract human respiratory signals around 0.2~0.4 Hz. Park et al.[9] placed PVDF piezoelectric film sensors beneath the mattress at the back of subjects to detect respiratory signals.

2.2

Millimeter wave radar sleep respiratory detection

FMCW radar emits continuous frequency-modulated pulses to measure the distance and velocity of targets[10]. The frequency of the pulse signal linearly increases over time, following a chirp modulation pattern expressed as Equation (2), where f_c is the starting frequency of the pulse signal, T_c is the chirp modulation repetition period, and B is the frequency bandwidth of the modulation.

The pulse signal and the target’s echo signal are mixed to get the IF signal. Millimeter wave radar measures the distance and speed of the target respectively through range FFT and Doppler frequency shift principle. Breath detection is achieved through the analysis of micro-Doppler effects. The frequency of human respiratory signals ranges from 0.1 to 0.5 Hz, with amplitudes on the body surface varying from 1 to 12 mm.

the relationship between thoracic displacement changes between consecutive pulses and phase variation is expressed by Equation (3), where ΔR represents thoracic displacement during successive pulse period, and λ denotes the wavelength of the pulse signal. By continuously analyzing the phase dynamics of linearly modulated pulses at the target’s Range Bin, digital signal processing algorithms can extract respiratory signals from the phase signals[11][12]. In 2020, Emmi Turppa et al.[13] studied the monitoring of vital signs in various sleep scenarios using FMCW radar, validating the practical application of millimeter-wave radar in sleep respiratory detection.

3.1

Experimental preparation

Experimental Subjects: The experiment involved 8 subjects aged between 20 and 31, all in healthy condition without any respiratory or cardiovascular diseases. Basic information, including age, gender, and physical condition, was collected from each subject prior to the experiment.

Measurement Devices: The millimeter-wave radar measurement method utilized the AWR1642BOOST module from Texas Instruments (TI) along with the DCA1000 EVM data capture board. The collected millimeter-wave signals were converted by an ADC and transmitted to a PC for data processing. For the piezoelectric mattress measurement method, a PSP-type flexible piezoelectric film sensor from was used. The signals were filtered and then transmitted to the PC via an RS232 serial port.

Data Collection Procedure: The millimeter-wave radar was positioned 70 cm directly above the subject’s chest, while the piezoelectric mattress was placed under the subject’s back, ensuring that the piezoelectric film sensor and the millimeterwave radar were approximately aligned along the same vertical axis, as shown in Figure 1. The subjects were instructed to lie in three positions sequentially: supine, prone, and lateral, with a 1-minute interval between each position. Data from both measurement methods were collected simultaneously and repeated three times. The millimeter-wave radar data were saved as bin files, and the piezoelectric sensor data were saved as csv files.

Figure 1.

Experimental scene diagram

3.1.1

Supine experiment

Figure 2 illustrates the supine position of the subject lying on the bed with the head aligned flat against the headboard. The millimeter-wave radar and the piezoelectric mattress synchronously sense and capture minute displacements on the thoracic surface caused by respiratory. The millimeter-wave radar collects a total of 1200 frames of data at a sampling rate of 20 Hz. On the host computer, the data undergo preprocessing, MTI (Moving Target Indication), arctangent phase extraction, phase unwrapping, and phase differencing to obtain the phase difference signal. Subsequently, the signal is filtered with a sliding average filter to remove pulse noise and normalized using a 4th-order Butterworth band-pass filter with a cutoff frequency of 0.1 Hz and stop-band cutoff frequency of 0.5 Hz, resulting in the normalized time-domain waveform of the respiratory signal collected by the millimeter-wave radar. The piezoelectric sensor collects a total of 30,000 frames of data at a sampling rate of 500 Hz. The data undergo the same Butterworth band-pass filtering and normalization process to yield the normalized time-domain waveform of the respiratory signal collected by the piezoelectric sensor. As shown in Figure 3, the blue line represents the respiratory signal from the millimeter-wave radar, and the red line represents the respiratory signal from the piezoelectric film sensor.

Figure 2.

Supine diagram

Figure 3.

Comparison of supine respiratory signals

The paper separately computes the mean and standard deviation of the respiratory signal peaks within 1-minute interval for both detection datasets, serving as indicators of the average intensity and stability of the respiratory signals. Figure 4 presents the mean and standard deviation of the respiratory signal peaks acquired under the supine sleeping posture. From this, preliminary observations suggest that the ratio of the standard deviation to the mean (Coefficient of Variation, CV) for the piezoelectric method is smaller compared to that of the millimeter-wave radar method.

Figure 4.

Mean and standard supine respiratory signals

3.1.2

Prone experiment

Similarly to the supine experiment, Figure 5 depicts the mean and standard deviation of the respiratory signal peaks acquired from all subjects in the prone sleeping position. Compared to the aforementioned experiments, it is evident that the peak values of respiratory signals collected via the millimeter-wave radar method are significantly higher in the prone posture than in the supine posture. In contrast, the peak values of respiratory signals obtained using the piezoelectric method do not show any notable differences between the two postures.

Figure 5.

Mean and standard prone respiratory signals

3.1.3

Lateral lying experiment

Similar to the supine experiment, Figure 6 depicts the subject in a lateral sleeping position. It was found that the peak value of the millimeter-wave radar in the lateral sleeping position is significantly lower than in other sleeping positions, whereas the piezoelectric method shows no significant difference.

Figure 6.

Mean and standard lateral respiratory signals

3.2

Analysis of sleep respiratory detection experiment results

Here are the summaries for the peak respiratory signals in the three sleeping positions (supine, prone, and lateral) presented in Figure 7. The experimental results indicate that the peak respiratory signal collected by the millimeter-wave radar in the lateral lying position is less than 0.02 dB, showing significant attenuation compared to other sleeping positions, ranked from highest to lowest intensity as prone, supine, and lateral. On the other hand, the peak respiratory signal collected by the piezoelectric method ranges between 22 to 28 mV, with no significant changes in respiratory intensity observed during changes in sleeping positions. Additionally, the CV (Coefficient of Variation) for the peak respiratory signal waveforms collected by both sensors was calculated and is shown in Figure 8. The results indicate that the piezoelectric sensor provides more stable respiratory signals in the supine and lateral lying positions compared to the millimeter-wave radar. Conversely, the millimeter-wave radar shows more stable respiratory signals in the prone lying position compared to the piezoelectric sensor.

Figure 7.

Comparison of respiratory signal means

Figure 8.

Comparison of coefficient of variation