ARCHIVES

Bowel sound (BS) results from the intestinal activity and can be observed via auscultation. Many studies have been conducted for earlier, harmless, and practical detection of bowel related diseases from BS. Typically, BS is observed using microphones but because of its sudden changing character, silent periods (QP) can last a very long time in the recordings, single bursts (SB) appear vaguely, and multiple bursts (MB) can be confused with external noise. In this study, a new method in BS analysis is proposed to attenuate noise, detect BS, determine the number of occurrences, and classify differentiate BS or Non-BS signals. A common vector approach (CVA) method was used for the detection and classification of BSs. This study aims to show that the orthonormal projection matrix (Q) can be used as a reflection tool in all recordings independent of the training set, without consulting medical professionals. The results show that, in the training, the BS recognition rates were 96,1%, non-BS recognition rates were83,3%, on the other hand, Single burst (SB) and Multiple bursts (MB) recognition rates were 80,9% and 15,4%, respectively. In the tests with using unused abdominal records, 87.5% of the single bursts (SB), 35,7% of multiple bursts (MB) and 84,3% non-BS signals were recognized. The ratios of SB, MB, QP and non-BS signal parts in the signal, were respectively 2,3%, 0,3%, 92,9% and 4,5%.