Artificial intelligence can recognize human emotions by voice: Where can it be used?

The artificial intelligence model can recognize emotions such as anxiety and fear in people's voices. The developer of the model hopes that it will help telephone operators in crisis centers in preventing suicide.

Assessing the emotional state of those who call crisis lines is important for identifying and preventing suicide. Voice can convey information about a person's mental and emotional state, but it is difficult for an operator to assess the risk of suicide, especially in conditions of extreme emotional instability of callers.

As Newatlas.com tells, a solution to this problem was proposed by Alaa Nfissi, a graduate student at Concordia University in Montreal, who trained an artificial intelligence model to recognize speech emotions. Typically, this assessment was carried out by psychologists, which required a lot of time and experience, but the deep learning model was able to effectively recognize emotions.

To train the model, we used a database of real recordings of calls to crisis lines and recordings of actors conveying certain emotions. The results showed that the model accurately recognizes fear/anxiety (82% correct answers), sadness (77%), anger (72%) and neutral state (78%).

Alaa Nfissi is confident that the model he developed can be used as an auxiliary tool for working on crisis lines, helping operators assess the condition of their interlocutors in real time and make appropriate decisions. This, he believes, can significantly help prevent suicide.