Speech bandwidth extension based on speech phonetic content and speaker vocal tract shape estimation

In this paper, we introduce a new speech bandwidth extension (BWE) algorithm which involves phonetic and speaker dependent estimation of the high-band part of the spectral envelope. Speech phoneme information is extracted by using a hidden Markov model. Speaker vocal tract shape information corresponding to the wideband signal is extracted by a codebook search. The proposed method allows better estimation of high-band formant frequencies, especially for voiced sounds, and better estimation of spectral envelope gain, especially for unvoiced sounds. Postprocessing of the estimated vocal tract shape allows artifacts reduction in cases of erroneous estimation of speech phoneme or vocal tract shape. We present experimental results that demonstrate improved wideband quality for different speech sounds in comparison to other BWE methods.