Direct coding of a class of non-stationary signals based on mixed transforms

A mixed Fourier-Walsh signal representation is proposed for transform coding. Basis functions from two transforms are used, namely, the discrete Fourier transform (DFT) and the Walsh-Hadamard transform (WHT). An iterative algorithm is used to implement the mixed representation. A finite number of DFT components is used to represent the selective (narrowband) dominant spectrum of speech on a frame-by-frame basis. The whitened (broadband) residual signal is represented using a small number of Walsh components. The parameters of the mixed representation are coded directly by coding the DFT and WHT components. Results are given for frame-by-frame and packet coding at rates ranging from 16 kb/s to 9.6 kb/s.<>