Abstract
With the aging population, telecare systems are more and more popular since patients can be cared by wearing a sensor rather than being restricted in hospitals. However, with various signals to be monitored continuously, power consumption has become an important issue in this field. Feature extraction is the main concern in our design. Due to the lower complexity and the better time-frequency analysis, discrete wavelet transform (DWT) algorithm has been applied to analyze biomedical signals such as Electrocardiography (ECG) and Electroencephalography (EEG) here. In ECG signal processing, we can use DWT algorithm to remove unusable features from original signals, and extract R-R interval sequences from the reconstructed waveform for diagnosing heart-related diseases. In EEG signal processing, we also can use the algorithm based on DWT to observe frequency domain features in Parkinson's disease (PD). Moreover, we have proposed a configurable wavelet processor with feature extraction circuit according to the proposed algorithm. It can be integrated in the sensor to extract signal features for more efficient biomedical applications. At the DWT computer of the processor, we have developed a structure which can not only let us configure filter coefficients by external microcontroller, but also save about 50% of the hardware complexity. We have tested our design on FPGA emulation and have implemented it with TSMC 0.18 µm technology. The total core area is 1.15 mm2, the operating voltage is 1.8 V, the operating clock frequency is 360 Hz, and the power consumption is 0.52 µW. Compared with sending all raw ECG data, our design saves as much as 99.5% power while only detecting and sending R-R interval sequences in ECG application.