This paper presents an alternative mode selector based on neural networks to improve the low-complexity AMR WB+ standard audio coder especially at low bit rates. The AMR-WB+ audio coder is a multi-mode coder using both time-domain and frequency-domain modes. In low complexity operation, the standard encoder determines the coding mode on a frame-by-frame basis by essentially applying thresholding to parameters extracted from the input signal and using a logic which favors time-domain modes. The mode selector proposed in this paper reduces this bias, and achieves a mode decision which is closer to the full complexity encoder. This results in measurable quality improvements, in both objective and subjective assessments.
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.