Novel method for the efficient individualization of binaural audio signals
Seiichiro Shoji, Anthony I. Tew · The Journal of the Acoustical Society of America · 2006
A novel coding technique suitable for the transmission of binaural audio signals with concurrent multiple sound sources is proposed in this paper. Existing algorithms typically suffer from audible artefacts, poor localization, or excessive bandwidth. Especially with conventional intrasource coding methods, bandwidth is dependent on the number of sources and becomes greater as the number of sources increases. The proposed binaural transformation (BX) encoder synthesizes binaural signals using generic HRTFs and transmits them to the decoder along with side information. The bandwidth of the side information is independent of the number of sources. Using locally stored HRTFs matching those of the listener, the decoder turns the generic binaural signals into personalized ones. We present two types of BX codec: type A and type B. Psychoacoustic experiments based on Recommendation ITU-R BS.1116-1 were used to compare signals processed by these codecs with the corresponding unprocessed signals. For a variety of source materials (speech, vocals, guitars, and percussion) we show that high mean opinion scores are achievable using the type A codec when its parameters are set appropriately.