Ego noise reduction for hose-shaped rescue robot combining independent low-rank matrix analysis and noise cancellation

Abstract

In this paper, we present an ego noise reduction method for a hose-shaped rescue robot developed for search and rescue operations in large-scale disasters such as a massive earthquake. It can enter narrow and dark places covered with rubble in a disaster site and is used to search for disaster victims by capturing their voices with its microphone array. However, ego noises, such as vibration or fricative sounds, are mixed with the voices, and it is difficult to differentiate them from a call for help from a disaster victim. To solve this problem, we here propose a two-step noise reduction method as follows: (1) the estimation of both speech and ego noise signals from an observed multichannel signal by multichannel nonnegative matrix factorization (NMF) with the rank-1 spatial constraint, which was proposed by Kitamura et al., and (2) the application of noise cancellation to the estimated speech signal using the noise reference. Our evaluations show that this approach is effective for suppressing ego noise.