PPG Noise Reduction based on Adaptive Frequency Suppression using Discrete Fourier Transform for Portable Home Monitoring Applications

Photo-plethysmogram (PPG) is the most widely monitored signal for the personalized home monitoring applications and wearable devices. PPG data recorded by such devices are often corrupted with different artifacts which significantly affects the diagnosis accuracy. Use of complex noise reduction techniques based on advanced signal processing tools is unsuitable for implementation in such devices due to their limited computational ability and power requirement. This paper proposes a noble and simple algorithm for noise reduction from PPG data. It is based on the removal of the noise frequencies by suppression of the corresponding Discrete Fourier Transform coefficients. The frequency components to be suppressed are chosen adaptively based on the individual signal characteristics. The performance of the algorithm is extensively tested with data collected from the MIMIC II Waveform database and with real data collected using the Biopac MP-150 data acquisition system for different types of added noises. Apart from displaying significant noise reduction ability as quantified by the SNR Improvement factor, the algorithm also well maintains the signal morphology. Moreover, the simplicity and ease of implementation of the proposed technique make it highly suitable for implementation in the portable home monitoring applications.