SPEECH SELECTIVE AUTOMATIC GAIN CONTROL

An automatic gain control circuit uses a speech recognizer to obtain smooth automatic gain control. An analog audio input signal is converted to a digital signal by an analog-to-digital converter and delayed by a delay circuit. A frame power (or alternatively, rectified peak amplitude) detector determines the power of each frame (or alternatively, the rectified peak amplitude) of the audio input signal, after applied to the A/D converter. A linear-to-log converter converts those values to a logarithmic form (for gain control over a broad range of values). A detected speech smoothing circuit smooths the variation in the values determined by the frame power (or peak amplitude) detector. A summer subtracts the output of the detected speech smoothing means from a fixed reference level, and thus obtains an error signal from the desired reference. A gain smoothing circuit smooths the resulting error signal (which is the logarithmically-shaped gain signal). A logarithm-to-linear converter converts the logarithmic gain signal to a linear form; and a multiplier multiplies the input signal by this smoothed gain. In accordance with the invention, a speech recognizer determines whether the audio input signal represents speech. An output of the speech recognizer is used to enable the detected speech smoothing circuit and the gain smoothing means when the audio input signal represents speech. Thus AGC is not used until it is required (i.e., when speech is present).