An efficient method is proposed for frame loss concealment within the advanced audio coding (AAC) decoder, which can effectively mitigate the adverse impact of frame loss on reconstruction quality. The spectral information of the lost frame is first estimated in the modified discrete cosine transform (MDCT) domain via the known frame interpolation approach. The interpolated MDCT coefficients are then further refined by magnitude scaling and sign correction, which are differently designed for tonal and noise components of the source signal: In noise-like spectral bins, shaped-noise insertion technique is employed to adjust the interpolated coefficients, while coefficients in tone-dominant bins are refined by magnitude scaling and novel sign correction techniques so as to optimize the fit of the corresponding time reconstruction with available partial signal information from neighboring frames. Subjective quality evaluations demonstrate that the proposed method achieves significant quality improvement over the shaped-noise insertion method adopted in commercial AAC decoders.
Click to purchase paper as a non-member or login as an AES member. If your company or school subscribes to the E-Library then switch to the institutional version. If you are not an AES member and would like to subscribe to the E-Library then Join the AES!
This paper costs $33 for non-members and is free for AES members and E-Library subscribers.