Heart Sound Localization in Respiratory Signals

Respiratory sounds have since long been important indicators of a person’s health and disease. Prior to the advances in computing and DSP, physicians relied on their hearing to detect abnormal signs in the respiratory sounds of their patients. Advances in computer technology and the ability of signal processing algorithms to detect symptoms and derive characteristic features of the respiratory sounds has led to significant research in the field of respiratory sound analysis. In this paper, we see that heart sounds form a major source of interference in respiratory signals. We look at two methods to localize heart sounds in respiratory signals: an entropy-based approach and a method based on multiscale products.

Respiratory sounds are recorded over the chest wall and digitized using microphones or accelerometers. The frequency range of interest lies between 50-2500Hz. Hence, the sampling frequency should be at least 5kHz. The peak of lung sounds is in frequencies below 100Hz. Lung sound energy drops off sharply between 100Hz and 200Hz, but it can still be detected at or above 800Hz with sensitive microphones. Heart sounds have a significant component in frequencies between 60Hz and 100Hz. Hence, as seen, heart sounds and lung sounds have an overlap in their peak frequency range. During data acquisition of respiratory sounds, heart sounds form an inevitable source of interference, and in order to process the respiratory signal for abnormalities, mitigating the effect of heart sounds is an important pre-processing stage in respiratory sound analysis.