Convolutional Neural Network (CNN) Based Speech-Emotion Recognition

Abstract

Speech is considered as the widest and most natural medium of communication. Speech can convey a plethora of information regarding one's mental, behavioral, emotional traits. Besides, speech-emotion recognition related work can aid in averting cyber crimes. Research on speech-emotion recognition exploiting concurrent machine learning techniques has been on the peak for some time. Numerous techniques like Recurrent Neural Network (RNN), Deep Neural Network (DNN), spectral feature extraction and many more have been applied on different datasets. This paper presents a unique Convolutional Neural Network (CNN) based speech-emotion recognition system. A model is developed and fed with raw speech from specific dataset for training, classification and testing purposes with the help of high end GPU. Finally, it comes out with a convincing accuracy of 83.61% which is better compared to any other similar task on this dataset by a large margin. This work will be influential in developing conversational and social robots and allocating all the nuances of their sentiments.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 72 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.