Despite of being a subjective branch of study, emotion analysis can be simulated by Machine Learning algorithms that are trained so that, through cataloged audio and video databases, they can recognize patterns in these media that could be related to their emotion . Neural Network Algorithms (branch of Machine Learning) provided to act on the recognition of these emotions, with a focus only on audio, and are known as Speech Emotion Recognition (SER), generally obtaining unequal averages of referring results such as recognition of emotions between the bases. This research work focuses on implementing an SER algorithm, based on current works that have achieved good performances, applying a Data Augmentation method called Slide Window, proposed by this work, which aims to increase the averages of recognition in selected databases. As a result, this work managed to achieve an evolution of recognition of 11.95% on the EMO-DB base, 22.76% on SAVEE and 18.82% on RAVEDESS.
|
