Wavelet Decomposition with Machine Learning for PPG Signal Quality Detection

As technology has advanced, the trend of using cloud connected smart phone based digital healthcare solutions is growing. The wireless wearable and smart phones based physiological signals measurements are essential parts of digital healthcare, offering significant advantages to many stakeholders regarding social, geographical, consultancy delays, and economic constraints. However such measurements can be affected by the artifacts and noise which can degrade the performance of post classifiers. It is crucial for signal processing and artificial intelligence (AI) assistive health monitoring applications to estimate the signal quality before transmitting or further processing it. This strategy could result in increased robustness and dependability. In this regard, a new hybridization of machine learning and wavelet decomposition (WD) based feature extraction is suggested for an automated identification of the smartphone sensors based acquired Photoplethysmography (PPG) signals. The feature extraction not only renders a notable data dimension reduction without sacrificing important information but also enhances the performance of post classifiers in terms of precision and latency. The categorization is carried out using the robust machine learning (ML) classifiers. The applicability is tested using a real multi-class and multi-subjects smartphone acquisition based PPG dataset. The devised solution achieves a highest classification accuracy of 98.00% with a notable dimension reduction.