Single Channel Spectrum-based Speech Enhancement Using Neural Networks

Abstract: The ability to communicate is fundamental to form a relationship, and it is anecessity for a well-functioning society. Since a major part of our daily communicationtakes place orally, the ability to perceive speech is important. However,it is not always as easy to perceive the message, especially when the level of backgroundnoise partially masks the speech. For a person with hearing impairment,the situation gets even worse.The impact of background noise is also challenging other domains, and one ofthose is regarding virtual assistants, which have recently become more commondue to the technological advancements. Since virtual assistants have allowedus to interact with our technological devices in our daily lives, the dependencythat they work becomes more critical. This dependency especially holds whenwe are required to interact with them by speech. Still, in both suggested cases,background noise remains an issue to some degree. Hence, the possibility toreduce the noise influence is likely to have a significant role in how our societydevelops.In this report, we evaluate the possibility of reducing background noise.To do it, we proposed a new neural network architecture which is based on theprinciples of extreme learning machine. Considering that this report works withspectrum-based speech, appropriate constraints to ensure non-negativity in ouroptimization problem has been carried out. Moreover, different configurationsapplied to the architecture have been observed, which includes unprocessed vs.pre-processed features, masking filter, and stacking several single architecturelayers.The results show that the proposed architecture with the unprocessed, noisyspeech, input performs better than an input pre-processed by a well-knownmethod. Another finding observed was that relaxation in constraint yieldedbetter performance of a noisy speech than based on a non-negative convexconstrainedsolution.