A novel complexity reduction technique using visibility relationship and perpendicular distance recursive refinement for physiological signals

Abstract

Signal simplification is a processing technique that reduces the number of samples in a signal. It has been employed in various applications and methods while handling huge amounts of data. One well-known method is the Douglas-Peucker (DP) algorithm which performs signal simplification using an appropriate tolerance value to determine whether to retain or remove a given sample point. That would mean the performance of the DP algorithm is sensitive to the selection of the tolerance value. In this paper, we introduce a new signal simplification method insensitive to parameter dependence changes. We first construct a connectivity-based visibility relationship matrix to find the most important points in the signal. Then, we use the degree threshold value to construct a degree matrix determining key anchors of the simplification process that preserve the essential features of the signal. This signal is simplified by measuring the perpendicular distances of the intermediate points from line segments defined by these key points. The proposed technique was tested on three simulated signal models and an electroencephalography (EEG) signal. Our results obtained are visually and quantitatively compared in terms of root mean square error (RMSE), R2, number of simplified points, and compression ratio with the DP algorithm. The results indicate that the proposed method is robust to parameter changes and provides better simplification than the DP algorithm. In the future, we plan to validate our algorithm with a huge database.