A Robust and Computationally Efficient Speech/Music Discriminator
A New method for discriminating between speech and music signals is introduced. The strategy is based on the extraction of four features, whose values are combined linearly into a unique parameter. This parameter is used to distinguish between the two kinds of signals. The method has achieved an accuracy superior to 99%, even for severely degraded and noisy signals. Moreover, the low dimensionality of the feature space, together with a very simple information-merging technique, has resulted in a remarkable robustness to new situations. The low computational complexity of the method makes it appropriate for applications that demand real-time operation. Finally excellent resolution for the segmentation of audio streams is achieved by manipulating the analyzed data properly.
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 temporarily free for AES members.