Apparatus for decoding a DPCM encoded signal

Claim 1

A voice decoding apparatus, comprising: a receiving demodulator for receiving and demodulating a signal, which is a multiplex-modulated signal of an ADPCM-coded voice signal and a voice activity detection flag indicating either a voice-active duration or a voice-nonactive duration and is transmitted for only the voice-active duration; a demultiplexer for demultiplexing the demodulated signal from the receiving demodulator into a coded voice signal and the voice activity detection flag; a controller for outputting the voice activity detection flag from the demultiplexer as it is and outputting a reset signal for each time when the voice activity detection flag changes from the voice-active duration to the voice-nonactive duration and vice versa; an adaptive differential PCM decoder for decoding, in a voice-active duration, the coded voice signal from the demultiplexer with reference to a prediction coefficient, which raise up from the zero state reset by the reset signal, and for decoding, in a voice-nonactive duration, random code words generated from inside circuitry thereof, with reference to a prediction coefficient of background noise in the voice-active duration just preceding to the current voice-nonactive duration, thereby to generate pseudo-background noise; and a prediction coefficient holder for receiving the voice activity detection flag from the controller, for receiving, in the voice-active duration, the prediction coefficient from the adaptive differential PCM decoder to calculate an average value thereof for each frame and to store the average value calculated after updating and for applying the average value of the prediction coefficient to the adaptive differential PCM decoder, and for temporarily interrupting, during the voice-nonactive duration, said updating and applying the average value of the prediction coefficient of the voice-active duration just preceding to the current voice-nonactive duration to the adaptive differential PCM decoder as said prediction coefficient of background noise; characterized by an auxialiary controller, provided between the controller and the prediction coefficient holder, for outputting the voice activity detection flag from the controller intact for respective lasting durations of the voice-active duration and the voice-nonactive duration, and for interrupting the updating of the prediction coefficient by applying the voice activity detection flag to the prediction coefficient holder for a period of time during which the prediction coefficient reset to the zero state in the adaptive differential PCM decoder raises up to a value substantially equal to the prediction coefficient of a voice-nonactive duration just preceding to the current voice-active duration, when the voice-nonactive duration changes to the current voice-active duration, thereby reducing a range of change in the tone quality of the pseudo-background noise, which is caused when the voice activity detection flag received indicates erroneously the voice-active duration for a short time in the voice-nonactive duration.