In the next Machine Learning Group seminar, we have a talk by visiting researcher Hazrat Ali.
Date & Time: Wednesday 2nd April, 1-2pm
Location: Room AG01 (College Building)
Title: Hybrid Features Combination for Audio Data Classification
Abstract: In this work, we present an approach to combine learning features and hand-crafted features for Audio data classification. In particular, we use Restricted Boltzmann Machine (RBM) for useful features extraction from audio data and combine them with MFCC features. These features can be used for gender classification, speaker classification tasks. We present our initial results for gender classification using features extracted from RBM as well as for traditional Mel Frequency Cepstral Coefficients. We suggest that combining the RBM features with traditional features can result into better performance.