Implementation of Mel-Frequency Cepstral Coefficient as Feature Extraction using K-Nearest Neighbor for Emotion Detection Based on Voice Intonation

Purpose: To determine emotions based on voice intonation by implementing MFCC as a feature extraction method and KNN an emotion detection method.Design/methodology/approach: In this study, the data used was downloaded from several video podcasts YouTube. Some of methods in study are pitch shifting for augmentation, audio data, basic statistics taking mean, median, min, max, standard deviation each coefficient, Min max scaler normalization process classification.Findings/result: Because testing is carried out separately gender, there two classification models. male model, highest accuracy obtained at 88.8% included good fit model. female 92.5%, but model unable to correctly classify new data. This condition called overfitting. After testing, cause because augmentation one tone women solve problem training size being too small not containing enough samples accurately represent all possible input values.Originality/value/state art: The research has never been previous studies downloading Youtube then processed until ready be research.