Design And Algorithm Study Of Wearable System For Pulse Signal Detection

With the increasing of aging society and the development of wearable devices, wearable medical devices have become a hot field. The wearable medical devices used to detect the pulse signal have gradually been applied to personal care. However, there are many limitations in commercial products, such as inconvenience to wear, no real-time monitoring in motion, etc.In order to solve above problems, a pulse signal processing algorithm was proposed for wearable watch, which focused on the pulse characteristics detection algorithm based on Haar Wavelet and supplemented with the motion artifact reduction algorithm based on adaptive Least Mean Square(LMS) filter. With the advantage of simple mathematical operation, the division of Haar transform can be replaces by the shifts, reducing the requirement of MCU processor capability and system resource. While the motion artifact reduction algorithm used data collected from the tri-axial accelerometer as reference noise source and the motion state detection algorithm used them as motion signal. According to the motion state, the algorithm determined the need to use the motion artifact reduction algorithm in advance for eliminating the motion artifact. With the pulse characteristics detection algorithm based on Haar Wavelet, the pulse signal can be distinguished from power line interference, baseline drift and high-frequency noise. The modulus maximum pairs of the wavelet transforms can be identified, which lead to detect the heart rate of pulse signal.Based on above algorithms, a wearable, low-power, low-cost health watch was designed, which can detect heart rate in still or motion. By reflective photoelectric detection method, pulse signals passed through trans-impedance amplification, secondary amplification and low pass filter circuit based on the second-order Chebyshev, finally went through the analog to digital conversion to MCU processor for digital signal processing. The selection of circuit chip, circuit design and circuit control program were demanded for low-power dissipation and high integration density. In order to improve the battery life, the LEDs were modulated by the narrow Pulse-Width Modulation(PWM) with a duty ratio of 25%, lowering the power consumption to 2.5mW.By using above algorithms, the detection rate of the pulse characteristics detection algorithm is above 99.18% for the PhysioNet database, the average measured deviation compared with products of Heart Rate Chest Strap on the market is about 5%. The motion artifact reduction algorithm can reduce the motion artifact from the pulse signal with the SNR value of-7dB. The heart rate can be detected further by the pulse characteristics detection algorithm after reducing the motion artifact.