SPEECH EMOTION RECOGNITION WITH MULTISCALE AREA ATTENTION AND DATA AUGMENTATION
SKIVE PROJECTS SPEECH EMOTION RECOGNITION WITH MULTISCALE AREA ATTENTION AND DATA AUGMENTATION In Speech Emotion Recognition (SER), emotional characteristics often appear in diverse forms of energy patterns in spectrograms. Typical attention neural network classifiers of SER are usually optimized on a fixed attention granularity. Convolution Neural Networks We designed an attention-based convolution neural network with 5 convolution layers, an attention layer, and a fully connected layer. The result is fed into four consecutive convolution layers and generates an 80-channel representation. Then the attention layer attends to the representation and sends the outputs to the fully connected layer for classification. Batch normalization is applied after each convolution layer. CONTACT US -Click here