Design of robust and efficient event-synchronous adaptive filter for real-time biomedical signal analysis

Abstract

In this dissertation, simple and efficient event synchronous delta function based adaptive filter, which is computationally superior having conditional weight update loops, has been designed for real-time cancellation of interference in several biomedical signals. A train of delta-like waves as a sequence has been used as a reference signal of the proposed adaptive filter, and each delta wave is synchronized to the primary events that need to be traced. Due to its structural characteristics, the algorithm is named as the event synchronous adaptive filter (ESAF). To evaluate the performance of the ESAF, simulation signals with ECG and white noise have applied to the ESAF and conventional FIR filter (600th) and analyzed the detection of ECG. MSE (Mean Square Error) has been used as analysis parameter. The ESAF present s much better performance than the conventional FIR filter.The ESAF is very simple in complexity but robust to impulsive interferences contained in the input signals. Its schemes mostly employ simple Boolean and rare multiplication operations and achieve considerable speed up over the other adaptive algorithm based implementations. Therefore the proposed implementation is suitable for applications such as wearable, wireless and ambulatory biomedical systems, where large signal to noise ratios with less computational complexity are required. In this dissertation, several primary biomedical signals ( ECG, EEG, FECG, BP) have been selected, implemented and analyzed for availability of the ESAF in biomedical signal analysis. The main components of applications are signal modeling and processing into four particular biomedical applications.In each four applications, results using the ESAF, the classical adaptive methods and nonlinear signal processing method (ICA) have been compared and analyzed for the signal-to-noise ratio (SNR) improvement achieved with each algorithm. Results studied in four applications showed that the proposed realization provided better performance and considerable advantages compared to existing conventional methods in terms of signal to noise ratio and computational complexity. Hence a successful solution based on the ESAF can be implemented excellently using low-power and low-cost microcontrollers which are alternative to commercial embedded medical systems for real-time biomedical signal monitoring instead of high-cost DSP (Digital Signal Processor) and FPGA. I foresee that the proposed method can be used as a powerful technique in the real time biomedical applications.