8.3 ADAPTED WP ALGORITHMS

The “best basis” methodologies [Coif92] [Wick94] for adapting the WP tree structure to signal properties are typically formulated in terms of Shannon entropy [Shan48] and other perceptually blind statistical measures. For a given WP tree, related research directed towards optimal filter selection [Hedg97] [Hedg98a] [Hedg98b] has also emphasized optimization of statistical rather than perceptual properties. The questions of perceptually motivated filter selection and tree construction are central to successful application of WP analysis in audio coding algorithms. The WP tree structure determines the time and frequency resolution of the transform and therefore also creates a particular tiling of the time-frequency plane. Several WP audio algorithms [Sinh93b] [Dobs97] have successfully employed time-invariant WP tree structures that mimic the ear's critical band frequency resolution properties. In some cases, however, a more efficient perceptual bit allocation is possible with a signal-specific time-frequency tiling that tracks the shape of the time-varying masking threshold. Some examples are described next.

8.3.1 DWPT Coder with Globally Adapted Daubechies Analysis Wavelet

Sinha and Tewfik developed a variable-rate wavelet-based coding scheme for which they reported nearly transparent coding of CD-quality audio at 48–64 kb/s [Sinh93a] [Sinh93b]. The encoder (Figure 8.5) exploits redundancy using a VQ scheme and irrelevancy using a wavelet packet (WP) signal ...

Get Audio Signal Processing and Coding now with the O’Reilly learning platform.

O’Reilly members experience books, live events, courses curated by job role, and more from O’Reilly and nearly 200 top publishers.