2008 Volume 9 Issue 1 Pages 51-57
Bowel sound (BS) is a biological audio signal stemming from peristaltic activity of intestines. We investigate the effects of a kind of foods on BS as the first attempt to develop a design method of custom-made foods by BS analysis. In this paper, a new procedure to detect BS by multidimensional signal processing is developed, leading to high detection accuracy of them, since an automatic BS detection system is absolutely required in order to process enormous quantities of BS data. The procedure, which involves our previous works, consists of independent component analysis, wavelet filtering, pattern matching in frequency domain, and threshold processing. In particular, using the new wavelet filter boosts the robustness and the new threshold processing decreases the miss detection ratio, compared to our previous works. Through a performance evaluation test of the new method, high detection ratio was achieved, which averaged 99%. Then, the false detection ratio was less than 2% and the additional detection ratio was more than 270%.