DSP board-based pulmonary symptoms detection using sound processing through acoustical stethoscope

In this paper, an automatic wheeze and crackle detection system is developed. Lung sounds are recorded in a wave file using an acoustical stethoscope amplifier circuit connected to the Blackfin 537 DSP board. The spectrogram is generated from the recorded wave file using Fast Fourier Transfor (FFT). The spectrogram image is passed through a bilateral filter and a limiter to increase the contrast and isolate the higher components. Image processing is used to detect wheezes in the image. Katz-Sevcik fractal dimension (KSFD) is then used to detect crackles. KSFD measures the complexity of a signal. Applying the concepts and techniques used in this study, the system was able to correctly detect normal sounds with an accuracy of 75%, wheezes with an accuracy of 62.5% and crackles with an accuracy of 91.67%