Abstract
In this thesis, a recognition system of congestive heart failure (CHF) based on heart rate variability is proposed. Therein, we attach great importance to the discussion of short- and long-term analysis. For the long-term analysis, we accessed 4-hour data; for the short-term analysis, we retrieved 5-minute data. Furthermore, all the ECG data is derived from the 24-hour recording of the clinical trial in National Taiwan University Hospital. This system is primarily divided into three parts: feature extraction, feature selection, and classification. To extract appropriate features, not only the basic linear features but also the non-linear methods such as the multiscale entropy, and detrended fluctuation analysis, are applied in our research. To further find out the correlation between the feature and disease, we employed the statistical method to verify whether there is a significant difference of features between CHF and Control or not. After excluding the features that are not sufficiently effective, we chose the sequential forward selection (SFS) to search for the best feature set for the short- and long-term analysis respectively. Afterwards, the support vector machine (SVM) is used to look for the best support hyperplanes to hive off the unclear group and then classify with the three groups. The comparisons of short- and long-term analysis are also discussed in this thesis. According to the leave-one-out method, we verified the performance of this recognition system. Finally, the recognition accuracy for long-term analysis is up to 100% and for short-term analysis attained 95.83%.