When encoding audio at very low bitrates using a transform coding technique like AAC, the encoder may not be able to encode the full audio bandwidth with good quality. For this reason, bandwidth extension techniques have been proposed, like Intelligent Gap Filling, that copy low-frequency spectral bands to high frequencies and adjust their properties by a set of parameters. Although such methods provide a significant quality improvement, the selected bands and their interaction with neighboring bands may cause artifacts mainly due to misaligned tonal components. In this contribution, a new method is proposed to generate high-frequency content for Intelligent Gap Filling. It is based on creating non-linear distortions in the time-domain decoded core signal. For tonal signals, this will create a harmonic spectrum with potentially better tonal alignment. As a non-linearity, sign- and RMS-preserving squaring of the input signal between pairs of consecutive zero crossings is used. A perceptually controlled dynamic selection mechanism is investigated to chose between the alternatives of no extension, copy-up and non-linear distortion. Listening tests show that for some tonal items the dynamic selection obtains an improvement in audio quality.
(See document for exact affiliation information.)
https://aes2.org/publications/elibrary-page/?id=23093
(645KB)
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 Join the AES. If you need to check your member status, login to the Member Portal.
