Harmonic excitation LPC (HE-LPC) speech coding at 2.3 kb/s

This paper presents an algorithm for encoding a speech signal at 2.3 kbit/s based on a uniform harmonic modeling of the excitation signal. The algorithm uses the robust pitch detection and efficient voicing analysis to split the LPC excitation into two bands. The lower band is related to the voiced parts of speech, while the upper band represents unvoiced speech. A fixed phase spectrum from a voiced segment generated by a male speaker is added into the uniform harmonic modeling of the excitation signal. This kind of fixed phase reduced the buzz effectively and produced soft natural speech. A short-term post-filter is utilized at the decoder to enhance the quality of synthesized speech. Subjective testing in Chinese showed that the 2.3 kbit/s HE-LPC coder performance is better than that of the federal standard 2.4 kbit/s MELP coder.