Biological Sound Sensor Robust to Air Conduction Noise

In this study, wearable biological sound sensor system is proposed for measuring pulse and respiratory rate. First, polyurethane elastomer as filler is adopted to match the specific acoustic impedance with the human body and reduce air conduction sound. Furthermore, HPSS signal processing is applied to separate the acquired biological signals into vascular sounds and respiratory sounds. Moreover, to remove air conduction noise from respiratory sounds, a novel respiratory sound extraction method using convolutive non-negative matrix factorization is proposed. In the proposed CNMF-based respiratory sound extraction method, the system can learn the characteristics of the frequency change of the respiratory sound, and thus can extract only the respiratory sound. The effectiveness of the proposed new respiratory sound extraction method is verified by the accuracy of the signal-to-noise ratio and respiratory rate measurement. Finally, the proposed system will enable us to measure the body in a noisy environment, such as in daily life, and will enable early detection of respiratory and cardiovascular diseases and telemedicine.