METHOD AND APPARATUS FOR MULTI-SENSORY SPEECH ENHANCEMENT

摘要

Embodiments of the invention include a method of determining an estimate of a clean speech value. The method includes receiving an alternative sensor signal from a sensor other than an air conduction microphone, receiving a noisy air conduction microphone signal from an air conduction microphone, identifying which frequency of a group of candidate frequencies is a pitch frequency for a speech signal based on the alternative sensor signal, using the pitch frequency to decompose the noisy air conduction microphone signal into a harmonic component and a residual component by modeling the harmonic component as a sum of sinusoids that are harmonically related to the pitch, and using the harmonic and residual components to estimate the clean speech value by determining a weighted sum of the harmonic and residual components. The clean speech value represents a noise-reduced signal having reduced noise relative to the noisy air conduction microphone signal.