Acoustic environment identification using blind de-reverberation

Apart from the actual signal content, a digital audio recording has various artifacts and distortions due to multiple reflection of sound and other activities in the environment which produce background noise. One such parameter which results in artifacts in the audio recording is the acoustic reverberation. This reverberation depends on the room geometry. This paper describes a methodology to extract the reverberation from the audio recording which can assist in identifying the environment in which the recording was done. Mel-Frequency Cepstral Coefficients are used as features while the multi-class Support vector machine (SVM) is used for classification of different environments. The Hand-Clap impulsive model was used to find the room impulse response of each environment. The performance was evaluated with six different environment and with five recordings in each environment.