Audio Signal Processing and Coding

6.1 INTRODUCTION

Audio codecs typically use a time-frequency analysis block to extract a set of parameters that is amenable to quantization. The tool most commonly employed for this mapping is a filter bank of bandpass filters. The filter bank divides the signal spectrum into frequency subbands and generates a time-indexed series of coefficients representing the frequency-localized signal power within each band. By providing explicit information about the distribution of signal and hence masking power over the time-frequency plane, the filter bank plays an essential role in the identification of perceptual irrelevancies. Additionally, the time-frequency parameters generated by the filter bank provide a signal mapping that is conveniently manipulated to shape the coding distortion. On the other hand, by decomposing the signal into its constituent frequency components, the filter bank also assists in the reduction of statistical redundancies.

This chapter provides a perspective on filter-bank design and other techniques of particular importance in audio coding. The chapter is organized as follows. Sections 6.2 and 6.3 introduce filter-bank design issues for audio coding. Sections 6.4 through 6.7 review specific filter-bank methodologies found in audio codecs, namely, the two-band quadrature mirror filter (QMF), the M-band tree-structured QMF, the M-band pseudo-QMF bank, and the M-band Modified Discrete Cosine Transform (MDCT). The ‘MP3’ or MPEG-1, Layer III audio codec pseudo-QMF ...