Improved spectral subtraction speech enhancement algorithm in cloud terminal speech interaction

To use the resources and services of cloud manufacturing through cloud terminal, the natural man-machine interaction mode was researched. Thus the resources of cloud manufacturing could be used efficiently by multiple channels. Speech interaction is the most natural interactive channel for human, and the natural language signal must be preprocessed in practical application to obtain clear speech signal from noisy speech. An improved spectral subtraction speech enhancement algorithm based on auditory masking effect was proposed. A new speech endpoint detection method was introduced to overcome the music noise influence and to extract the pure speech. The simulation results verified the effectiveness of the proposed algorithm.