Speech Emotion Recognition Using Audio Matching

It has become popular for people to share their opinions about products on TikTok and YouTube. Automatic sentiment extraction a particular product can assist users in making buying decisions. For videos languages such as Spanish, the tone of voice be used determine sentiments, since translation is often unknown. In this paper, we propose novel algorithm classify sentiments speech presence environmental noise. Traditional models rely pretrained audio feature extractors humans that do not generalize well across different accents. leverage vector space emotional concepts where words with similar meanings have same prefix. example, starting ‘con’ or ‘ab’ signify absence hence negative sentiments. Augmentations are way amplify training data during classification. However, some augmentations may result loss accuracy. Hence, new metric based eigenvalues select best augmentations. We evaluate proposed approach emotions YouTube outperform baselines range 10–20%. Each neuron learns pronunciations emotions. also use model birds from recordings city.