EEG average power analog feature extraction
Georgios Ntinopoulos
Abstract
The subject of this thesis is the design of a low-power biomedical analog system that will be suitable for extracting the average power of each brain rhythm-frequency band. This system will consist of the appropriate bandpass filters to separate the EEG signal to its constituent frequency bands, namely delta, theta, alpha, beta and gamma. Each band also has the necessary amplifiers to boost the signal as well as rectifiers with lowpass filters at the output to average the rectified signal. The system will be implemented using MOS transistors biased in the subthreshold region. The supply is at Vdd=-Vss=450mV and the bias currents are in the order of nA, this allows the implementation of the large time constants necessary for the small biomedical frequencies to be realized. The goal of this thesis is to develop an analog system that will calculate some features in the analog domain and in effect eliminate the need to calculate them in the digital-domain part of the system ,thus alleviating extra power consumption. The circuits were designed and implemented using the Cadence software with TSMC 90nm CMOS Process technology. Finally, the necessary calculations for the filters were accomplished using MATLAB.