This paper reports on some improvements to the audio-coding method transform-domain weighted interleave vector quantization (TwinVQ). In the encoder, the frequency-domain coefficients are flattened at two stages before vector quantization is applied. The paper also introduces Bark-scale envelope coding. There are three frame modes to change time/frequency discrimination. According to the subjective quality tests, reproduction quality of 40 kb/s TwinVQ exceeded that of MPEG : Layer III.
This paper costs $33 for non-members and is free for AES members and E-Library subscribers.