Spectracentrogram: A Time-Frequency Distribution for Signal Processing Applications

In this paper, we propose a spectracentrogram model for time-frequency analysis of major signal-power concentration. The proposed model is derived from a traditional time-frequency representation (TFR) scheme, spectrogram. The algorithm consists of two major operations: performing short time Fourier transform (STFT) for spectrogram, and computation of center-of-gravity (CoG) for a power spectrum. Compared to the STFT spectrogram, the proposed spectracentrogram shows a single sharper spectral curve, which represents the resultant direction of signal-power corresponding to the varying CoGs. The performance of the proposed model is illustrated on three computer simulated signals and three practical cardiac signals such as electrocardiogram (ECG), seismocardiogram (SCG), and photoplethysmogram (PPG). Qualitative analysis of the results shows the good performance of the proposed method and its extendibility for different signal processing applications.