Abstract:
"
Deep learning is a subset of machine learning which facilitates the machines to work
independently to simulate some particular tasks without explicit programming or without
any human intervention. The researchers are using deep learning algorithms in different
domains such as healthcare, robotics, education, and many more because of the
tremendous features that deep learning comprise of such as different tuning capabilities,
high accurate predictions, etc. Out of all the domains, speech emotion recognition is one
of the main areas that the researchers applying deep learning concepts to build more
robust applications such as to conduct sentiment analysis, usage of psychological
applications, online medical consultation systems, and more.
Moreover, speech emotion recognition is using in call centers to alleviate various
problems related to customers as well as call agents. Reducing the customer waiting time
of urgent customers is one of the major problems that researchers are trying to resolve.
Therefore, they have been devoted several years to find out a suitable solution and finally
focused on deep learning despite the effort which was taken based on machine learning
and other latest technologies.
When considering the latest efforts which are undertaken to reduce the customer waiting
time, those studies reported excellent results but still can be improved. In order to address
this issue differently, this research is mainly focused on the hybrid acoustic features in
the human voice and a CNN-based architecture. After going through many test scenarios,
the work carried out in this study achieved tremendous success for the average number
of voice recordings. The system reported 7.232 seconds for 40 voice calls as the overall
processing time and this is a good achievement when compared with the existing
systems. Furthermore, the implemented EMOLIZER system will be very useful to
identify the urgent customers, especially during this COVID-19 pandemic."
