Signal filtering of oximeter based on LMS optimization algorithm

Non-invasive monitoring of blood oxygen saturation is essential in clinical settings. However, the photoplethysmography (PPG) signal is highly susceptible to interference from environmental light frequencies and motion artifacts. These factors can lead to signal distortion, reduced smoothness, and baseline drift, thereby compromising the accuracy and reliability of the monitoring process. Traditional optical filtering schemes are limited by their large size, poor flexibility, and inability to adapt to dynamic interference. This paper proposes an adaptive notch filter based on LMS algorithm, which constructs a composite interference model of power frequency harmonics including 50Hz, 100Hz, 150Hz components and low-frequency motion artifacts covering the frequency range of 0.05 to 0.5Hz. Orthogonal sine-cosine reference signal matrix is used to achieve multi-component synchronous estimation, and time-varying frequency tracking mechanism and stray light compensation strategy are introduced to dynamically optimize weight updates. MATLAB simulation verification shows that this method significantly improves the signal-to-noise ratio by 14.71dB to 30.47dB (an increase of 15.76dB) compared to traditional low-pass filtering schemes, reduces the mean square error by 88.5% to 0.0028, and achieves a correlation coefficient of 0.994 between the output waveform and the pure PPG signal. Time frequency domain analysis further confirms its ability to deeply suppress power frequency interference and harmonic components, while effectively eliminating low-frequency motion artifacts that are difficult to handle with traditional methods. This scheme provides a feasible signal processing path for the development of high anti-interference and low-power oximeters.