This book describes the study and development of several methods for the bandwidth extension of speech signals. The proposed methods include the extraction of speech–model based parameters like cepstral coefficients, auto–regressive parameter (all–pole–filter coefficients), line spectral pairs (LSP) or mel–frequency cepstral coefficients (MFCC). These parameters and their properties are briefly presented. Besides these parametric representations of the spectral envelope some other scalar speech features like the zero–crossing rate, gradient index, pitch analysis, local kurosis, spectral centroid and energy based features are introduced.

The book also focuses on the performance of these methods, using the objective spectral distortion measures. A significance analysis is also drawn forthe subjective evaluation as well as a rank correlation between the subjective results and the respective objective distortion measures to evaluate the significance of these objective measures. Finally, the most promising method is implemented in a real–time environment using appropriate hardware.