A hybrid transform method for analysis/synthesis of speech
Andreas Spanias · 2002
A mixed Fourier/Walsh transform scheme for speech coding is proposed. A set of harmonically structured frequency components is used to represent narrowband components of speech. The broadband residual is characterized by a small number of sequency components. The frequencies and sequences of the transform components are determined by sampling the short-time Fourier and Walsh transforms, respectively. The magnitudes and phases of the Fourier components and the amplitudes of the Walsh components are determined using an iterative algorithm which is based on the Gauss-Seidel method. A vector quantization (VQ) scheme is developed to encode the frequency and the sequency components. Results and subjective evaluations are given for speech coding at 4.0 kbit/s.>