In-Ear Microphone Speech Data Recognition using HMMs

Many applications requiring detection and identification of speech while in high noise environments can be commonly found in factory, automobile, aircraft or other settings. In such conditions, collecting speech at other locations than the mouth may lead to speech of better quality than can be obtained at the mouth. Our study focuses on a COTS foam-incased in-ear microphone device well suited for multiple users and various environments. We present results obtained with a basic DHMM recognizer implemented on a seven isolated words vocabulary