Integration of a Robust Clustering Algorithm and Multiple Sensors to Achieve the Noise Cancellation of Heart Rate

Many wearable devices are commonly used to measure vital signs; heart rate is the most frequently monitored physiological information. Due to the weak signal strength of heartbeats, capturing these signals presents a significant technical challenge. Therefore, this paper adopts a non-invasive wearable device for detecting heartbeats. A wearable device is constructed using three polyvinylidene fluoride (PVDF) piezoelectric film sensors, placed at the three endpoints of an equilateral triangle with a side length of 3 cm, and positioned near the heart to detect heartbeat signals. The multiple sensors in this wearable device utilize the vibration signals to cause deformation in the PVDF piezoelectric film, generating voltage amplitude to represent the magnitude of the vibrations. Since the sensors are very sensitive to detect vibration signals, both physiological signals and surrounding noise are detected when the heart beats, resulting in a low signal-to-noise ratio (SNR) for heart rate signals and significantly increasing the chances of incorrect heart rate interpretation. In this paper, to improve the SNR, not only is hardware circuit design employed to amplify the signals and eliminate high-frequency noise using a low-pass filter, but Agglomerative Hierarchical Clustering algorithm (AHC) is also used to group and classify the datasets by the three sensors. Irregular signals that deviate from the clusters are treated as noise, thereby eliminating noise from the signals and improving the quality of the physiological heart rate signals. According to the experimental results, the SNR of the heart rate signal after noise cancellation can be increased by 7 dB, and the accuracy of heart rate signal recognition can reach 98.46%.