Audio Signal Processing and Coding

12.8 ITU-R BS.1387 AND ITU-T P.861: STANDARDS FOR PERCEPTUAL QUALITY MEASUREMENT

In 1995, the Radio Communication Sector of the International Telecommunications Union (ITU-R) convened Task Group (TG) 10/4 to select a perceptual measurement scheme capable of predicting the results of subjective listening tests in the form of an objective difference grade (ODG). Six proposals were submitted to TG 10/4 in 1995, including the distortion index (DIX) from TU-Berlin [Thie96], the noise-to-mask ratio (NMR) from FhG-IIS Erlangen [Bran87a], the perceptual audio quality measure (PAQM) from KPN [Beer92a], the perceptual evaluation (PERCEVAL) from CRC Canada [Pail92], the perceptual objective measure (POM) from CCETT, France [Colo95], and the Toolbox (TB) from IRT Munich. During initial TG 10/4 tests in 1996, three variants of each proposal were required to predict the results of an unknown listening test. A lengthy comparison and verification process ensued (e.g., [ITUR94a] compares NMR, PAQM, and PERCEVAL). Ultimately, the six TG 10/4 proponents in conjunction with Opticom and Deutsche Telekom Berkom collaborated to design a single system combining the most successful features from each individual methodology. This effort resulted in the recently adopted ITU-R Recommendation BS.1387 [ITUR98], which specifies two versions of the measurement scheme known as perceptual evaluation of audio quality (PEAQ). The basic, low-complexity version is based on an FFT filter bank, much like the PAQM, PERCEVAL, ...