ABSTRACT
Strange attractor can be constructed from time series data such as heart sound. In the areas of the recognition and diagnosis of abnormal heart sounds, signal presentation method is very useful because good features can be detected from good presentation. This paper examines efficiency in diagnosing abnormal heart sounds of the two different methods for constructing attractor. Nine different heart sounds from typical clinical conditions were used for this study. The first method was constructing attractors using original heart sounds, and the second was modifying the original sounds by autocorrelation and they were then applied to the orignal sounds as to cross correlation checks. Attractors could be constructed using signals generated by these methods, and values of fractal dimensions would then be calculated which has been a well known method to measure characteristics of attractors. The results showed that the second method appeared to provide more efficient way to correctly classify abnormal heart sounds.
Subject(s)
Diagnosis , Fractals , Heart Sounds , HeartABSTRACT
Strange attractor can be constructed from time series data such as heart sound. In the area of the recognition and diagnosis problem, signal presentation method is very important because good features can be detected from good presentation. This paper discusses a way to extract a cycle from strange attractor and introduce new attractor construction method using autocorrelation value of the heart rate. The result shows well-formed attractor and good ability for extraction features. Largest Lyapunov Exponent is used to check whether the attractors provide distinguish abilities among different types of heart rate. The result shows good points that can be applied to some areas of human signal processing.