<![CDATA[Photoplethysmography (PPG) is a non-invasive technique for measuring various physiological parameters, including blood glucose levels. However, PPG signals are often affected by noise and artefacts that reduce the accuracy of analysis and prediction. Therefore, an effective noise filtering method is needed to make the signal quality and ready for feature extraction for blood glucose estimation. This study offers a solution to the problem of noise in PPG signals through the application of appropriate pre-processing methods. This study aims to select quality PPG signals through three pre-processing methods: detrend, smoothing, and 0.5-5 Hz bandpass filter. The effectiveness of the three methods was evaluated through ADF test to measure the stationarity of the signal, frequency spectrum analysis to observe the distribution of frequency components, and SNR test to assess the signal to noise ratio. Based on the analysis of 67 data samples, the p-value <0.05 was obtained, indicating that the signal has reached a stationary condition. In addition, the average test statistic of men is higher than that of women, indicating that men's signals are more stationary after detrend. Meanwhile, 36 samples (54%) had SNR ≥ 20 dB indicating that more than half of the data were of good enough quality for further analysis. The results show that multi-stage pre-processing improves the quality of PPG signals, validated through quantitative tests of stationarity and SNR values. Thus, the preprocessed and improved quality PPG signals are considered feasible for use in the development of estimation models for various physiological parameters, including blood glucose levels.]]>
