Research project: Independent component analysis in the automated detection of evoked potentials from multichannel recording
As presented at the Signal Processing and Control Group Away Day, June 2012.
As presented at the Signal Processing and Control Group Away Day, June 2012.
Noise reduction in multichannel evoked potential data, SNR improvement, is the aim of the current part of this project. Independent component analysis (ICA) is one of the techniques which has become more popular in noise reduction of multichannel signal recording. In this work, a new method of noise reduction based on automated independent component selection is presented and tested using simulated data. Then the algorithm is applied to 68-channel EEG recording of Auditory Late Response (ALR) and the result is compared with existing methods (e.g. coherent averaging). The result of comparing SNR calculated by Fsp from different methods shows that SNR is improved considerably using the new method, i.e. the mean and standard deviation of SNR across the channels in the new method is found 9.43 and 5.5 respectively, whereas it was 1.38 and 1.3 using the averaging method. Moreover, SNR in the best channel is found 17.74 which is improved by factor of 3.7 in comparison with averaging method that gave 4.76. The next step will be to apply the method to a new larger dataset.