Modified Spectral Subtraction Method to Reduce Noise in sEMG Respiratory Muscle Signals

Characteristics of sEMG respiratory signals during breath holding indicate person's physical condition in a non-invasive manner. Baseline and interference noise typically strongly contaminate these sEMG signals and make detection and identification of important events difficult. In this paper we propose a modified spectral subtraction method for noise reduction that shows very promising results: it is able to strongly reduce unwanted signal components and still preserve features of interest. Comparison with reference methods shows an improved performance. At the same time, the proposed algorithm is computationally fast, making it suitable for real time implementation.