The nature of statistical learning theory
The nature of statistical learning theory
On the Learnability and Design of Output Codes for Multiclass Problems
COLT '00 Proceedings of the Thirteenth Annual Conference on Computational Learning Theory
Computer Methods and Programs in Biomedicine
| Hi-index | 0.00 |
An adaptive, support vector machine based ECG processing and compression method is presented in this study. The conventional pre-filtering algorithm is followed by a characteristic waves (QRS, T, P) localization. The regressive model parameters that describe the recognized waveformes are determined adaptively using general codebook information and patient specific data. The correct regocnition ratio of the QRS waves was above 99.9% using single channels from the MITBIH database files. The adaptive filter properly eliminates the perturbing noises such as 50/60 Hz power line or abrupt baseline shift or drift. The efficient signal coding algorithm can reduce the redundant data about 12 times. The good balance among proper signal quality for diagnosis and high compression rate is yielded by a support vector machine based system. The properly obtained wave locations and shapes, using a high compression rate, can form a solid base to improve the diagnosis performance in clinical environment.