REDUCTION OF INTERFERENCE IN KNEE SOUND SIGNALS BY ADAPTIVE FILTERING

The knee joint sound signal has been a subject of research with the aim of developing a non-invasive method for the diagnosis of cartilage abnormalities. Knee sounds in both sub-audible and audible ranges may be detected by a technique called vibroarthrography. Usually other sound signals from the active muscles interfere with the knee sounds. In order to achieve accurate knee sound analysis, the interfering signals should be cancelled. However, conventional linear filtering methods are inappropriate for this purpose since the interfering signals and the knee sound signals have spectra that overlap significantly. We propose an adaptive filtering method based on the least mean squares (LMS) criterion for the reduction of interference in the knee sound signals. Experimental results show that this method is able to reduce the interference signals significantly, while retaining the features of the desired knee sound signal.