transform-domain weighted interleave vector quantization (Audio) (twinVQ)

TwinVQ compression, also known as VQF, is a lossy audio compression developed at NTT Human Interface Laboratories in Japan based on the principle of vector quantization.

The TwinVQ process features a higher compression ratio and better audio quality compared to MP3, and is especially suitable for compressing music. The quality of TwinVQ at 96 kbit/s is comparable to that of MP3 at 128 kbit/s. Speech intelligibility is good at 8 kbit/s.

In TwinVQ compression, audio signals are sampled, the samples are considered as patterns of vectors, and compared to standard patterns. If a vector pattern corresponds to a stored standard pattern or comes very close to it, the index of the vector is transmitted. This pattern matching neglects interfering noise such as hiss or noise.

TwinVQ is part of MPEG-4 V.1.