Research on auto-identification method to the typical bowel sound signal

Bowel sound (BS) is an important parameter in clinical diagnosis and therapy. Auto-identifying the kind of BS has an important significance in helping the medical staff to analysis the status of the critical patients´ intestinal peristalsis and to select the most effective treatment. According to the BS signals´ characteristics, such as complicated noise, poor periodicity and strong randomization, a new processing method is proposed in this article. First of all, the signal´s energy and duration is taken as the separation indexes in extracting the BS segments to identify the BS occurrence counting per unite time. Secondly, adaptive noise reduction is used to the extracted signal to filter out the background noise. Thirdly, the characteristics of the purer BS signal is calculated to extract the classification rules. Finally the auto-identification of the BS signal can be achieved. The effectiveness of the method was verified by three kinds of BS signals (hyperfunction, normal and enteroparalysis) with the cooperation of a hospital intensive care unit (ICU) and the rules for three kinds of BS signals was given.