Spatial feature extraction from wrist pulse signals

Pulse diagnosis is an important diagnostic method in traditional Chinese medicine. However, it lacks objectivity. To standardize pulse diagnosis, pulse-taking platforms are required. This study implements a complete processing method for wrist pulse signals obtained from a pulse diagnosis instrument with a two-dimensional pressure sensor array to extract the spatial features. First, a zero-phase filter is adopted for acquiring the appropriate frequency band for bio-signals and removing noise. The filter is better to extract the pulse trend due to no phase shift distortion. Next, irregular pulses are removed prior to signal analysis. Then, percussion peaks are identified and the interval between them is used to calculate the pulse rate. Finally, a polynomial surface fitting method is used to compute pulse features, such as the peak, length, width, and surface curvature, from a visualized pulse, which is helpful for the study of pulse classification or even pulse diagnosis.