[A Wavelet-Based Time-Frequency Modeling Method and Its Application in Analysis of Local Field Potentials in Olfactory Bulb].

Qi Dong,Liang Hu,Liujing Zhuang,Jun Zhou,Ping Wang
DOI: https://doi.org/10.7507/1001-5515.20140089
2014-01-01
Abstract:The study of neuronal activity with low frequency has shown an increasing interest for its greater stability and reliability recent years. One challenge in analyzing this kind of activity is to find similarities and differences between signals efficiently and effectively. The traditional analysis methods, such as short-time Fourier transform, are easily obscured by background noises and often involve a large number of parameters. Therefore, this paper introduces a novel time-frequency analysis method based on wavelet transformation and half-ellipsoid modeling to extract instantaneous frequency and instantaneous phase information. This method overcomes some shortcomings of conventional time-frequency analysis. In this method, wavelet transformation is used to provide high-level representations of raw signals, and parsimonious half-ellipsoid models are used to extract changes in time domain and frequency domain of neural recordings. The method was validated to local field potentials (LFPs) of olfactory bulb of anesthetized rats during three different odor stimuli. The results suggested that this method could detect odor-relevant features from olfactory signals with large variability. The Odors then were classified with support vector machine (SVM) algorithm and the classification accuracy reached 79.4%.
What problem does this paper attempt to address?