Speaker Identification Enhancement Using Emotional Features

Speaker recognition is a broad field that encompasses many different tasks related to identifying speakers in audio recordings. Two specific sub-tasks that are often studied are speaker segmentation and speaker identification. These tasks typically involve analyzing acoustic features of the audio to determine who is speaking. However, one limitation of traditional speaker identification methods is that they can struggle when dealing with emotional conversations, as the acoustic features can change due to the emotions being expressed. To address this limitation, focuses on studying the effect of emotion on speaker identification by combining features of both the emotions and speakers. This approach has shown to improve identification accuracy, increasing it from 72% using speaker features alone to 75% when both emotion and speaker features are used.