abstract

Modern automated speech recognition (ASR) challenges are much more difficult than their predecessors' due to the apparent need from practical applications. Over time, the ASR system has improved to handle a wider range of challenges, including a larger vocabulary, more freedom to express oneself, more background noise, more diverse speech, and more languages. There has been a lot of recent activity in the area of speech emotion recognition (SER), which seeks to identify emotional states from signals in spoken language The most recent paper suggests using a deep garnitures algorithm to identify different emotions in a speaker's voice. When the system receives a voice signal, it analyses each individual feature using independent component analysis and the Fisher criteria. Both the computing time and the complexity of the system are drastically reduced by the suggested technique. Consequently, the difficulties of automated speech recognition and speech-emotion recognition are extensively covered in the thesis.