Speech Emotion Recognition Using Deep Feedforward Neural Network

Abstract

Speech emotion recognition (SER) is currently a research hotspot due to its challenging nature but bountiful future prospects. The objective of this research is to utilize Deep Neural Networks (DNNs) to recognize human speech emotion. First, the chosen speech feature Mel-frequency cepstral coefficient (MFCC) were extracted from raw audio data. Second, the speech features extracted were fed into the DNN to train the network. The trained network was then tested onto a set of labelled emotion speech audio and the recognition rate was evaluated. Based on the accuracy rate the MFCC, number of neurons and layers are adjusted for optimization. Moreover, a custom-made database is introduced and validated using the network optimized. The optimum configuration for SER is 13 MFCC, 12 neurons and 2 layers for 3 emotions and 25 MFCC, 21 neurons and 4 layers for 4 emotions, achieving a total recognition rate of 96.3% for 3 emotions and 97.1% for 4 emotions.<em><span style="font-size: 9pt; font-family: Arial, sans-serif;" lang="EN-MY">Speech emotion recognition (SER) is currently a research hotspot due to its challenging nature but bountiful future prospects. The objective of this research is to utilize Deep Neural Networks (DNNs) to recognize human speech emotion. First, the chosen speech feature Mel-frequency cepstral coefficient (MFCC) were extracted from raw audio data. Second, the speech features extracted were fed into the DNN to train the network. The trained network was then tested onto a set of labelled emotion speech audio and the recognition rate was evaluated. Based on the accuracy rate the MFCC, number of neurons and layers are adjusted for optimization. Moreover, a custom-made database is introduced and validated using the network optimized.</span></em><em><span style="font-size: 9pt; font-family: Arial, sans-serif;" lang="EN-MY">The optimum configuration for SER is 13 MFCC, 12 neurons and 2 layers for 3 emotions and 25 MFCC, 21 neurons and 4 layers for 4 emotions, achieving a total recognition rate of 96.3% for 3 emotions and 97.1% for 4 emotions.</span></em>

Development of video-based emotion recognition using deep learning with Google Colab

“…The scope for future improvements is very appealing in this field. Different multimodel deep learning techniques can be used along with different architectures to improve the performance parameters [20][21][22][23][24][25][26][27]. Apart from recognizing the emotions only, there can be further addition of intensity scale.…”

On the review of image and video-based depression detection using machine learning

“…They examined various feature sets for distinguishing depression from unconstrained speech and discovered loudness and intensity features to be the most discriminative. According to the technique used by [36], the Speech Emotion Recognition (SER) rate over five emotions, namely happy, angry, sad, fear, and neutral, achieved was with 91.7% accuracy by introducing Voice Activity Detection (VAD) in preprocessing. For creating a model based on speech indicators, it is vital to use a formalized dataset.…”

Improving spectrogram correlation filters with time-frequency reassignment for bio-acoustic signal classification

“…This has sparked interest in using automation to identify and classify animals from their vocalizations. This area of bio-acoustics signal analysis has been mostly concentrated on using techniques similar to those used for processing speech signals [2], [3]. Signal segmentation of bio-acoustic sounds is routinely performed in order to isolate syllables [4], [5].…”