EmoVoice — A Framework for Online Recognition of Emotions from Voice

Abstract

Abstract. We present EmoVoice, a framework for emotional speech corpus and classifier creation and for offline as well as real-time online speech emotion recognition. The framework is intended to be used by non-experts and therefore comes with an interface to create an own personal or application specific emotion recogniser. Furthermore, we describe some applications and prototypes that already use our framework to track online emotional user states from voice information.

FILTWAM and Voice Emotion Recognition

“…However we have lower values for an extended set of the emotions. This is in accordance with [28,34] in which we expect low accuracy for completely normal low-intensity emotions in online recognition and decreasing the classification accuracy with the number of emotional categories. This issue requires further investigation and improvement.…”

“…The NB classifier is very fast and appropriate for real-time emotion recognition. Our voice emotion recognition software supports speaker independent recognition approach, which is a general recognition system and therefore its accuracy is lower than the speaker dependent recognition approach that has been reported in [28].…”

“…Such devices not only offer opportunities for more natural interactions with the online game-based learning applications, but also offer ways of gathering affective user data during learning. There is a growing body of studies on online and automatic voice emotion recognition systems [28,29,30]. In one well-known study [28] the researchers have presented a framework for classifier creation for both offline and real-time voice emotion recognition with a specific interface for German speakers.…”

“…There is a growing body of studies on online and automatic voice emotion recognition systems [28,29,30]. In one well-known study [28] the researchers have presented a framework for classifier creation for both offline and real-time voice emotion recognition with a specific interface for German speakers. They allowed creating a dataset for joy, satisfaction, anger, and frustration emotions in the context of a training procedure for non-expert users.…”

Speaker-independent emotion recognition exploiting a psychologically-inspired binary cascade classification schema

“…At the second level, activation is applied. The reason for this choice is that when researchers limit themselves to a 2 dimensions emotional space instead of a 3 dimensions one, then this space is valence-activation [8] [14] [62]. Obviously, Fig.…”