No abstract

No abstract

No abstract

Typical speech enhancement algorithms operate on the short-time magnitude spectrum, while keeping the short-time phase spectrum unchanged for synthesis. We propose a novel approach where the noisy magnitude spectrum is recombined with a changed phase spectrum to produce a modified complex spectrum. During synthesis, the low energy components of the modified complex spectrum cancel out more than the high energy components, thus reducing background noise. Using objective speech quality measures, informal subjective listening tests and spectrogram analysis, we show that the proposed method results in improved speech quality.

In speech processing the short-time magnitude spectrum is believed to contain most of the information about speech intelligibility and it is normally computed using the short-time Fourier transform over 20-40 ms window duration. In this paper, we investigate the effect of the analysis window duration on speech intelligibility in a systematic way. For this purpose, both subjective and objective experiments are conducted. The subjective experiment is in a form of a consonant recognition task by human listeners, whereas the objective experiment is in a form of an automatic speech recognition (ASR) task. In our experiments various analysis window durations are investigated. For the subjective experiment we construct speech stimuli based purely on the short-time magnitude information. The results of the subjective experiment show that the analysis window duration of 15-35 ms is the optimum choice when speech is reconstructed from the short-time magnitude spectrum. Similar conclusions were made based on the results of the objective (ASR) experiment. The ASR results were found to have statistically significant correlation with the subjective intelligibility results.