An intelligent automatic level controller for speech signals: BELBIC

Abstract

Speech enhancement is a technology that is of high interest to both military and civilian voice communication developers and users. Automatic Gain Control — AGC; also called ALC is one among the level control techniques used for the purpose. An Automatic Level Controller (ALC) for speech signals embedded in additive noise requires Voice Activity Detection (VAD) to avoid noise amplification. The VAD is generally computationally intensive. This paper explores a less computational-intensive ALC, which uses a Brain Emotional Learning Based Intelligent Controller (BELBIC) for detecting voice activity. The controller tracks the energy variation of the signal by means of detecting rapid variations of the signal envelope. The segments that show a low variation of energy are detected as noise. The need for an Automatic Gain Controller (AGC) is to amplify speech segments to an intelligible sound level, while not amplifying noise only segments [1].