Poster board 63 - Tue 11/07, 16:45 - Hall X
Session 182 - Locomotion
Abstract A182.1, published in FENS Forum Abstracts, vol. 3, 2006.
Ref.: FENS Abstr., vol.3, A182.1, 2006
|Author(s)||Brittain J. (1), Halliday D. M. (1), Conway B. A. (2) & Nielsen J. B. (3, 4)|
|(1) Dept. Electronics, Univ. York, York, UK; (2) Bioeng. Unit, Univ. Strathclyde, Glasgow, UK; (3) Dept. Med. Phys., Panum Inst., Univ. Copenhagen, Copenhagen, Denmark; (4) Inst. Physical Exercise and Sports Science, Univ. Copenhagen, Copenhagen, Denmark|
|Title||Single-trial multi-wavelet coherence in locomotion studies.|
|Text||A method for single-trial wavelet coherence is presented which makes use of the class of generalised Morse wavelets and may be applied in a neurophysiological context. Such multi-wavelets provide a method of optimal time-frequency localisation while supporting the construction of bivariate parameters, such as coherence and phase. Coherence may be considered an intuitive measure of common synaptic input. Such parameters provide invaluable insight into the time-varying dynamics of neural connectivity and their changing oscillation patterns during physiological tasks. Due to the orthogonality inherent within the class of generalised Morse wavelets, the resultant coherence estimates will possess statistics (including confidence limits) which are easily described and may be considered constant across frequencies. Multi-wavelets derive from the multiple applications of wavelet functions to an underlying time series. Each application of a generalised Morse wavelet leads to the production of a wavelet-periodogram. By averaging across these wavelet-periodograms, smoothed wavelet spectra are constructed which may be considered statistically more reliable, with reduced bias and variance properties. Generalised Morse wavelets are optimally concentrated within a specified time-frequency region. The shape of this time-frequency region may be adjusted through the setting of two parameters to allow tuning of the analysis to the underlying physiological system. By applying more wavelets in the average, a more consistent spectral estimate may be constructed, however this is offset by a proportional expansion of the time-frequency localisation region. This presentation will illustrate the application of single trial multi-wavelet coherence analysis to simulated data and locomotion data.