Implementation of adaptive filters for biomedical applications

In biomedical signal acquisition like electrocardiography ECG or electroencephalography EEC one of the main problems is to separate the small input signals from noise and disturbances caused by the 50 Hz power supplies, high frequency interference and random body voltages. Different types of analogue and digital filters are used to remove the unwanted spectral parts. In most applications the filter bandwidth of those filtes are fixed and will not adapt to changing interference patterns. Adaptive filter techniques are required to overcome this problem. Different adaptive filter types have been analyzed. Finite Impulse Response (FIR) filters are prefered because of their better stability. An adaptive filter was implemented which suppresses known noise sources in an ECG application. Simulations were done with MATLAB and VHDL. The filter was coded in VHDL and tested on a FPGA. A 50 Hz interference on the ECG input signal was attenuated by 50 dB. The convergence time for the adaptive algorithm was less than 3 sec. The filter implementation needed 9500 equivalent gates and worked with 7.1 mW for a filter clock speed of 1.8 kHz.