Low Power FIR Filter Bank for EEG Processing Using Frequency-Response Masking Technique

Abstract

Different frequency bands in an electroencephalogram (EEG) signal contain different information. It is very helpful to divide an EEG signal by its sub-bands before applying further classification. FIR filter is one of the best choices for processing EEG signal because of its linear phase property. However, the implementation of an FIR filter requires more multipliers compared to its IIR counterpart. With frequency-response masking (FRM) technique, the multipliers needed to implement FIR filter can be reduced dramatically leading to a low power design. This paper proposes a filter bank structure for processing EEG signal based on the FRM technique. The design equations for all the sub-filters are derived and the condition for applying the proposed structure is given. A design example is included to illustrate the effectiveness of the proposed filter. It shows that the filter can fulfill the design objectives with 77% less multipliers comparing to the conventional FIR filter synthesizing technique.