A Noise Reduction Method for Biological Sound Using a Multi-Channel Winner Filter with Adaptive Margin

Abstract

The number of patients with cardiovascular and respiratory diseases such as COPD and atrial fibrillation is increasing in the world. These diseases are difficult to detect in their early stages because the initial symptoms are subtle. Currently, a device that can measure the cardiovascular and respiratory organs simultaneously and can be easily handled is required. Therefore, for early detection and early treatment, we had developed a biological sound measurement system that measures vascular and respiratory sounds simultaneously. The system obtains the biological sound by attaching the biological sound sensor on the head. After obtaining biological sounds, the system separates them into vascular and respiratory sounds by signal processing. In conventional systems, ambient noise affected the obtained sound negatively. In this paper, we propose a noise reduction method using a multi-channel Wiener filter. The system uses two inputs. The inputs are a main input that observes the target signal and noise, and a reference input that observes only the noise. A Wiener filter is designed from the input signal to reduce noise. In order to suppress the degradation of biological sound, a margin determined by the amplitude ratio of the input was added to the Wiener filter. The effect of noise reduction was evaluated by SNR. The SNR calculated from the validation experiment was 28.7 dB at maximum, and the system was confirmed that vascular and respiratory sounds can be obtained with low degradation. Moreover, the SNR required for automatic diagnosis was met, so the proposed system is expected to be applied for telemedicine.