Abstract
Although chest-mounted photoplethysmography (PPG) has proven useful in clinical applications, it nevertheless remains vulnerable to dynamic-range challenges. Because the chest is less well perfused than traditional PPG measurement sites, it yields weaker pulsatile signals that may be masked by large DC baselines. Some commercial sensors feature coarse built-in DC cancellation, but they inject considerable uncompensated artifacts. First, this paper presents a hardware–software codesign using a commercial sensor with a DC cancellation feature, which is overridden by smoothed, software-defined control inputs. A real-time calibrator estimates and cancels DC components using an exponentially weighted moving average. Second, a linear predictive coding-based artifact-elimination method is introduced. Third, a multi-channel data fusion method using a softmax combiner is developed to improve PPG signal quality. System performance was evaluated using experimental data from six human subjects in a seated laboratory environment. Results indicate that the methods remove DC offsets and step artifacts, producing waveforms that are similar to ground-truth references. Quantitative results also show that the proposed combiner achieved a signal-to-noise ratio of