A Survey for Emotion Recognition Based on Speech Signal

Abstract

There are many characteristics of human beings, such as fingerprints, DNA, and retinal pigmentation that are essential. A person's voice is unique to each individual. Humans use speech to communicate their thoughts and feelings. The process of determining one's mental state involves expressing one's basic emotions in words. A person's emotions play a significant part in his or her daily existence. In order to convey one's thoughts and feelings to others, it is essential to use this method. Emotions can be discerned from speech information because humans have a built-in ability to do so. The selection of an emotion recognition body (speech database), the identification of numerous variables connected to speech, and the selection of a suitable classification model are the main hurdles for emotion recognition. To identify emotions, an emotion identification system analyzes auditory structural elements of speech. The analysis is based on multiple research papers and includes an in-depth examination of the methodology and data sets. The study discovered that emotion detection is accomplished using four distinct methodologies: physiological signal recognition, facial expression recognition, a variety of speech signals, and text semantics in both objective and subjective databases such as JAFFE, CK+, Berlin emotional database, and SAVEE. In general, these techniques enable the identification of seven basic emotions. To determine the emotion, the audio expression for eight emotions (happy, angry, sad, depressed, bored, anxious, afraid, and apprehensive), all published research maintain an average level of accuracy. The major goal of this survey is to compare and contrast numerous previous survey methodologies, which are backed up by empirical evidence. This study covered signal collection processing, feature extraction, and signal classification, as well as the pros and downsides of each approach. It also goes over a number of strategies that may need to be tweaked at each step of speech emotion recognition.