Envelope analysis methods for tonality estimation

Abstract

In perceptual audio coders, the audio signal masks the quantization noise. The masking effectiveness depends on the degree of tonality/noisiness of the signal. Hence, in psychoacoustic models (PM) of perceptual coders, the level of the estimated masking thresholds can be adjusted by tonality estimation methods. This paper introduces three envelope analysis methods for tonality estimation: optimized amplitude modulation ratio (AM-R), auditory image correlation, and temporal envelope rate. The methods were implemented in a filter bank-based PM. In a subjective quality test, they were compared to each other and to another existing method, partial spectral flatness measure (PSFM). The PSFM and the AM-R were rated significantly higher than the other methods