الفهرس 
IntroductionOnly 14 pages are availabe for public view
 |  | from | 108 |  |  |
| | | from | 108 | | |
Abstract
Electroencephalograph (EEG) data compression is desirable for a number of reasons. Primarily, it decreases transmission time, archival storage space, and in portable systems, it decreases memory requirements or increases channels and bandwidth. One of the first goals was to automatically collect EEG?s data from hospitals or patients? homes, through low speed transmission media such as switched telephone lines or cellular telephones, with low cost hardware, without the need of a physician?s presence. The main purpose of the present study is to investigate thoroughly the performance of a number of the existing compression techniques. Specifically, the main objective is to compare the compression ratio achieved, and the difference between the original and reconstructed waveforms quantitatively for a number of selected techniques under the same circumstances in order to identify the technique that is able to compress the data without introducing information losses into the EEG signal. A variety of direct and transformation data compression techniques are investigated and compared. Besides the visual comparison, different criteria are used for performance evaluation: The compression ratio, the percent root mean square difference, root mean square error, the correlation coefficient, signal to noise ratio, and the maximum amplitude error. The results have shown that the direct data compression techniques give better performance than the transformation algorithms with the advantage of exact signal reconstruction, which is an essential requirement for physicians. The highest compression ratio obtained is 5.2245. The two techniques that give the highest values of compression ratio are the Huffman coding and repetition count compression techniques.