Identification of bowel sound signal with spectral entropy method

A bowel sound signal is very significant to clinical diagnosis of diseases, but it is a weak, non-stable and aperiodic signal generated under a strong noisy background. How to filter out the background noise effectively and provide real and clear signal for clinical diagnosis? This is one of the concerns in medical engineering field. This study proposes a detection method which is based on dynamic weighing threshold. With this method, the spectral entropy of every noise frame to be determined and the spectral entropy of noise in silent period are weighted to obtain new spectral entropy of noise which is used as the updated threshold. The spectral subtraction is used in the determination to increase signal-to-noise ratio for further decrease in noise interference. The simulation experiment results indicate that, compared to traditional method for spectral entropy endpoint detection, this method is more accurate in detecting the endpoint of bowel sounds in low signal-to-noise ratio condition.