SPEECH ENHANCEMENT BASED ON TURBO ITERATION

Hang Dong (1) , Hong Sun (1)

1 Signal Processing Lab

A Turbo iterative method for speech enhancement is proposed. The Kalman filter with the voice generation model and the wavelet threshold filter with the short-term spectrum are combined through Turbo iteration. Both filters work in rotation and each one takes some feedback information from the other filter as a priori condition. Our experiment results show that the Turbo iterative algorithm will converge within 10 iterations, and it achieves a good balance between the noise reduction and voice restoration.

Mots clés: Speech enhancement; Turbo iterative; voice distortion

Domaines: Traitement du signal et de l'image [eess.SP]; Traitement du signal et de l'image [eess.SP]

Hang Dong, Hong Sun. SPEECH ENHANCEMENT BASED ON TURBO ITERATION. 2009. ⟨hal-00385393⟩