Detection Of Pulse Oxygen Saturation In Strong Noise Background

The oxygen supply affects the metabolism in human body, and oxygen saturation is an important index of evaluating oxygen supply condition. Transmission pulse oximeter has been used widely in clinical and family health care which can realize non-invasive oxygen saturation detection. Traditional oximeter applies complex analogue electric circuits to realize signal processing, this leading to poor stability. Moreover, it is hard to eliminate the motion artifact. To solve these problems, this thesis proposes using software method to process pulse signal and extract its characteristic parameters. By implying morphology method, translation invariant lifting wavelet method, and empirical mode decomposition method, weak pulse signal submerged in noise is extracted. This improves detection accuracy. The main content on the paper is summarized as follows:Firstly, the hardware platform is designed and completed to detect dual-wavelength finger transmitted light. This platform is based on the ATmega8 MCU. The I/O ports of this MCU have nearly 20mA current driving capability which can activate the dual-wavelength light source directly. The internal integrated 10-bit analog-to-digital converter (ADC) in MCU also simplifies the circuit design.Secondly, we transmit the signal to computer in order to process pulse signal with software method. This paper combines the morphological filter with the translation invariant lifting wavelet method to remove baseline drift and high frequency noise in pulse wave. The comprehensive method combines their good merits. Morphological filter has small computational work and a high processing speed. Its performance in removing baseline drift is nearly perfect. Translation invariant lifting wavelet method has excellent performance in high-frequency noise removing. Comprehensive method can extract pulse wave signals excellently when the signal-to-noise ratio is quite low.Thirdly, the empirical mode decomposition method is used to eliminate the motion artifact and correct waveform distortion. Then the lifting wavelet modulus maxima method is used to realize feature point detection of the pretreated pulse signal. Compared to the traditional wavelet modulus maxima method, it achieves similar examination precision with less computation and high efficiency. Finally, pulse oxygen saturation is calculated on the basis of its definition and pulse signal's characteristic parameters. The result shows this system has achieved real time pulse oxygen saturation measurement and continuous monitoring of human body.A summation is made to generalize the work and deficiency in this paper. Suggestions are proposed for the further research and improvements.