Emotional Recognition of Netizens Voice Message on Social Media by using Deep Learning Neural Network Method

Communication activities between netizens represent the condition of curiosity to find out new information through social media such as Telegram, Line or WhatsApp. This has made the style response in the speaking process and represents the emotional condition. In this research, speech emotions were identified in Indonesian through voice notes on social media, especially WhatsApp. The speech is analyzed and classified to determine the conditional emotions of Indonesian netizens. The step started with composed the sound of audio dataset to the form of a file (mono.wav) and carried out by feature extraction using the Mel Frequency Cepstral Coefficient (MFCC) and classified by using the Deep Learning Neural Network. The classifier in this case is 1D Convolutional Neural Network with Swish Net architecture. System testing indicated a good performance on kernel 6 epoch 100, with output precision in the form of identifying 4 types of emotions Happy (93%), Angry (91%), Neutral (83%), and Sad (100^%).