In this paper, a distributed modular version of the Mallat algorithm is used as the finite impulse response (FIR) filter module of a field-programmable gate array (FPGA)-based discrete wavelet transform system for electroencephalogram (EEG) analysis. According to the authors, parallelization halves processing time while doubling the costs of computing units. The Daubechies wavelet function DB4 was used to convert floating-point coefficients into integers for FPGA use. The system has been simulated with Quartus II, and implemented on an Altera Cyclone III device. The testing setup fed the EEG data into the FPGA system, and the corresponding output data were displayed using either MATLAB or the EEG equipment itself. An example with both the original and the filtered EEG signals is presented, as well as a MATLAB chart with the extracted EEG rhythms.

The authors claim that their approach can minimize the loss of useful information, effectively extracting EEG rhythms that can help with the diagnosis of some brain diseases. They suggest that this could possibly lead to the development of portable EEG equipment. The approach to system implementation and the example presented by the authors seem very nice indeed. However, I would have preferred a better description of the applied tests and a more thorough discussion of the corresponding results.