A PYTHON LIBRARY FOR AUTOMATED WAVELET-BASED SIGNAL ANALYSIS AND NOVEL WAYS OF VISUALIZING CROSS-WAVELET PROPERTIES

Many movements, such as breathing, walking and chewing, are rhythmic, and thus require their neural circuits to generate rhythmic activity with regulated phase relationships of multiple components. Understanding how the nervous system generates the timing and patterns of activity are needed to understand how these circuits function and how they are affected by neurologic diseases.
To quantify rhythmic activities as in MATLAB-based software SpinalCore (Mor and Lev-Tov, 2007), we developed an open-source Python library that uses wavelet transformations and automates signal processing of multi-channel recordings over extended durations. For non-stationary signals, wavelet transforms achieve the optimal balance between resolutions in time and frequency domain (Torrence and Compo, 1998).
Our library incorporates automated segmentation of long-duration recordings, up/down-sampling, digital filtering, figure plotting, and export of results. In addition, we present a number of novel visualization methods: a multipaneled graph of scatter plots with each panel representing wave properties of one frequency band; a 2D contour plot of phase relationships across time and frequency; a scatter polar spectral plot that displays cross-wavelet phase, frequency, coherence and power, such that the complexities of cross-correlation between biological rhythms can be fully appreciated.
We also introduce a single frequency band analysis tool that isolates wavelet coefficients at specific frequencies of interest, enabling the analysis of cross-wave relationships between two frequency bands and shifts in cross-wavelet properties over hours of recording.
In summary, our Python library offers an automated pipeline for wavelet-transformation based signal analysis and provides novel ways for visualizing properties of rhythmic activities.