A low-complexity 4.6-kbit/s speech coder based on extremely truncated impulse response

This paper proposes a new low-complexity 4.6-kbit/s speech-coding method. This method uses an extremely truncated impulse response (ETIR) that allows the distance between the input speech and the synthesized speech to be calculated very quickly. Also, a new method of distance measurement to search for the optimum code vector eliminates the negative effect of the extreme truncation. Our coding experiments have shown that the proposed truncation method does not cause a significant decline in speech quality even if the number of FIR filter taps is lowered to three, and that our 4.6-kbit/s coder achieves a weighted signal-to-noise ratio that is only 0.3 dB lower than when using the conventional full-tap coder with conventional distance measurement.