Bandwidth Extension of Speech Using Perceptual Criteria

This book presents innovative algorithms for speech coding, focusing on bandwidth extension methods that improve efficiency and quality in Bandwidth Extension of Speech Using Perceptual Criteria.

The book Bandwidth Extension of Speech Using Perceptual Criteria delves into the intricacies of speech coding, specifically focusing on the bandwidth extension techniques employed within the International Telecommunication Union G.729.1 standard. This standard merges a narrowband bitstream with quantized high-band parameters, resulting in high-quality wideband speech. However, the authors identify opportunities to further reduce the additional bits needed for high-band representation, enhancing efficiency without compromising quality.

In addition to discussing the G.729.1 standard, Bandwidth Extension of Speech Using Perceptual Criteria introduces innovative bandwidth extension methods based on spectrum prediction. These methods offer a compelling alternative by eliminating the need for additional bits, although they may struggle in scenarios where the correlation between low and high bands is weak. The book presents two novel wideband speech coding algorithms that operate as wrappers around existing narrowband compression schemes, showcasing how to effectively integrate these techniques.

The first algorithm generates wideband speech using only transmitted high-band information, while the second employs a constrained minimum mean square error estimator that combines high-band envelope data with predictive schemes based on narrowband features. Both algorithms utilize advanced perceptual models centered on loudness to optimize quantization strategies for wideband recovery and synthesis. Through objective and subjective evaluations, the authors demonstrate that their proposed system achieves lower average bit rates while significantly enhancing speech quality compared to existing algorithms.