Speech emotion recognition with light gradient boosting decision trees machine

<p>Speech emotion recognition aims to identify the expressed in speech by analyzing audio signals. In this work, data augmentation is first performed on samples increase number of for better model learning. The are comprehensively encoded as frequency and temporal domain features. classification, a light gradient boosting machine leveraged. hyperparameter tuning determine optimal settings. As datasets imbalanced, class weights regulated be inversely proportional sample distribution where minority classes assigned higher weights. experimental results demonstrate that proposed method outshines state-of-the-art methods with 84.91% accuracy emo-DB dataset, 67.72% Ryerson audio-visual database emotional song (RAVDESS) 62.94% interactive dyadic motion capture (IEMOCAP) dataset.</p>