Audio Signal Processing and Coding

8.1 INTRODUCTION

Similar to the transform coders described in the previous chapter, subband coders also exploit signal redundancy and psychoacoustic irrelevancy in the frequency domain. The audible frequency spectrum (20 Hz–20 kHz) is divided into frequency subbands using a bank of bandpass filters. The output of each filter is then sampled and encoded. At the receiver, the signals are demultiplexed, decoded, demodulated, and then summed to reconstruct the signal. Audio subband coders realize coding gains by efficiently quantizing decimated output sequences from perfect reconstruction filter banks. Efficient quantization methods usually rely upon psychoacoustically controlled dynamic bit allocation rules that allocate bits to subbands in such a way that the reconstructed output signal is free of audible quantization noise or other artifacts. In a generic subband audio coder, the input signal is first split into several uniform or nonuniform subbands using some critically sampled, perfect reconstruction (or nearly perfect reconstruction) filter bank. Nonideal reconstruction properties in the presence of quantization noise are compensated for by utilizing subband filters that have good sidelobe attenuation. Then, decimated output sequences from the filter bank are normalized and quantized over short, 2–10 ms blocks. Psychoacoustic signal analysis is used to allocate an appropriate number of bits for the quantization of each subband. The usual approach is to allocate an adequate number ...