Quality controlled compression technique for Photoplethysmogram monitoring applications

Computerized acquisition and analysis of Photoplethysmogram (PPG) can provide vital information on various cardiovascular functions. In this paper, we introduce a quality controlled PPG compression technique using principal component analysis to keep the local and global distortion of the reconstructed data within pre-specified limits. To achieve this, optimum number of eigenvectors (EV) and principal components (PC) along with their quantization level (Q) were selected for compression. The technique was validated with pre-recorded 2 min and 10 min PPG data collected from 35 healthy volunteers (N) and 35 cardiovascular patients (CVP) under fully resting condition. At limiting PRD of 5% and absolute error of 8%, for N group (CVP group) an average compression ratio (CR), PRD and MAE of 5.65 (6.76), 2.41% (3.33%) and 6.1% (7.3%) were obtained respectively. Relaxing the control criteria up to clinically acceptable limits, a CR of 13.28 for N and 13.31 for CVP group were achieved. Most of the clinical features had an average deviation of less that 6% from their original values. The method yielded lower CR, PRDN and MAE with noisy (motion artifact and Gaussian noise) PPG. With two different stress levels, the compression parameters were found to vary. The reconstruction results were clinically validated by experts. A successful real-time data streaming in computer proved the usefulness of the technique for continuous monitoring applications.