Convolutional Neural Network Model for Detecting Anger from Sound Waves

Yiğit M., Kalaycı M. E., Turhan K.

15.TIP BİLİŞİMİ KONGRESİ, Trabzon, Türkiye, 30 - 31 Mayıs 2024, ss.212-222

  • Yayın Türü: Bildiri / Tam Metin Bildiri
  • Basıldığı Şehir: Trabzon
  • Basıldığı Ülke: Türkiye
  • Sayfa Sayıları: ss.212-222
  • Karadeniz Teknik Üniversitesi Adresli: Evet

Özet

The audio data also contains emotional content. Nowadays, it is aimed to develop a method that will help to determine the verbal and physical violence that prevents health workers from working in a safe environment. The model created to determine the verbal violence that healthcare workers will be exposed to due to the work environment. It will contribute to the prevention of possible violence by warning the security personnel at the time of violence with the help of machine learning from the audio data to be collected in the work environment. In this paper, they developed a model that detects anger from sound waves in order to provide a safe working environment for healthcare workers. The study used the CNN (Convolutional Neural Networks) method to analyze approximately 10-second-long sound fragments obtained from Turkish dubbed television series. The model was able to distinguish between angry and neutral voices with high accuracy and achieved a success rate of 75%. The results of the study show that higher success rates can be achieved when the data set is expanded, and that this model can be used for security purposes in the field of healthcare.