Eeg Compression Algorithm Based On Wavelet Transform

Electroencephalogram (EEG) signals are fundamental for the clinical diagnosis of patients with neural system diseases. There are four basic components which areδwave(0.5-3Hz), θwave (4-7Hz), α wave (8-12Hz), β wave (13-30Hz) in EEGsignals. In order to judge the different types of the diseases accurately and choose the suitable medicine and healing solution, people usually need to record a large amount of EEG data, which has brought the enormous pressure to storage and transmission. For this reason, the data should be compressed to maintain the main characteristics of EEG signals, which can achieve data reduction and solve the problems of storing, transmitting, searching and classifying etc.. There are several methods for EEG data compression, such as Huffman coding, DPCM/ADPCM, visiting distance coding, vector quantization, arithmetic coding, etc.. As we know, these algorithms are not specificly designed for the EEG signals, so the compression rate is not very high.In this thesis, a compression algorithm for EEG signal based on embedded zerotree coding (EZW) is proposed. EZW algorithm, which was proposed by Shapiro, is an effective image wavelet compression algorithm, and there is a similar zerotree structure in the one-dimensional signals' wavelet coefficient. In this paper an improved EZW algorithm, based on time-frequency characteristic of EEG signals, was discribed for one-dimensional EEG signals compression. When the EEG signals were compressed, the EEG signals were firstly s egmented. Then each segment of signals was transformed with wavelet and the wavelet coefficients were quantified with zerotree respectively. Finally Huffman coding was used to achieve more compression rate. We know that the classical wavelet transform is a lossy compression and inappropriate for lossless compression. So the transformed data must be truncated before being stored. Wavelet transform, whose wavelet basis is constructed based on lifting steps, is called the second generation of wavelet or integer wavelet. It can achieve integer-to-integer wavelet transform that can beapplied to lossless compression of signals. In order to realize the lossless compression of the EEG signals, a novel EEG compression algorithm is proposed by considering time-frequency characteristic of EEG signals. It includes predictive compression with delayed inputs, integer wavelet transform, zerotree coding and Huffman coding, which have achieved a better compression rate.A large amount of EEG signals were used to test this algorithm. The computer simulation experiments indicated that the algorithm of embedded zerotree coding we adopted is satisfying in compression performance and subjective observation. It is a very effective EEG signal compression method. And the proposed algorithm has excellent features of low complexity, high speed and high efficiency. High compression ratio can be achieved with high signal fidelity. It is especially suitable for the situation that real-time compressing was required such as tele-medicine.In this paper, some new algorithms and methods are proposed for the lossy and lossless compression of EEG signals. These algorithms and methods will be helpful to develop devices for EEG data storage and transmission.