Recording environment leaves its acoustic signature in the audio recording captured in it. For example, the persistence of sound, due to multiple reflections from various surfaces in a room, causes temporal and spectral smearing of the recorded sound. This distortion is referred to as audio reverberation time. The amount of reverberation depends on the geometry and composition of a recording location, the difference in the estimated acoustic signature can be used for recording environment identification. We describe a statistical framework based on maximum likelihood estimation to estimate acoustic signature from the audio recording and use it for automatic recording environment identification. To achieve these objectives, digital audio recording is analyzed first to estimate acoustic signature (in the form of reverberation time and variance of the background noise), and competitive neural network based clustering is then applied to the estimated acoustic signature for automatic recording location identification. We have also analyzed the impact of source-sensor directivity, microphone type, and learning rate of clustering algorithm on the identification accuracy of the proposed method.
Click to purchase paper as a non-member or login as an AES member. If your company or school subscribes to the E-Library then switch to the institutional version. If you are not an AES member and would like to subscribe to the E-Library then Join the AES!
This paper costs $33 for non-members and is free for AES members and E-Library subscribers.