CORROSION DETECTION IN CONCRETE USING PHASE-BASED SPEECH ENHANCEMENT AND RECOGNITION TECHNIQUES WITH AI FRAMEWORK

Abstract

Automatic Speech Recognition (ASR) is one of the oldest signal processing techniques which deals with detecting different types of speech including spoken words, emotions, disease, speaker identity, languages, music, and noise levels. Recently, these standard ASR methods have been adopted into the fields of acoustics, bioinformatics, wildlife, and environmental audio detection. One of the upcoming applications of ASR is in civil engineering which deals with the hammer sound test. This test is one of the non-destructive tests which is performed to detect the concrete quality and surface defects. This research problem has been taken in this paper and investigated with various phase-based audio and statistical features along with speech enhancement techniques to reduce unwanted noises. Phase of the audio signal plays a crucial role in extracting relevant audio information which is less explored in the audio signal processing. The proposed phase-based method has been tested with standard hammer sound datasets and other environmental sound datasets. It has been observed from the analysis of the simulated results that the proposed method has been demonstrating superior performance.