|
|
|
Scalable, Integrable to any application even in other fields and the accuracy is increased as compared to the previous models with similar approaches to 89% using more number of emotions. It predicted the emotion correctly in the testing database, except for two files when the emotions calm and neutral were included(that was for the CNN model). I’ve also tested with my own voice which is radically different from the test or the train data used, and my first language is Telugu, not English nor German and yet this system has recognized my emotions upto a good rate and giving an accuracy of 85%. For live recordings, it also has close predictions. The MLP, LSTM models gave state of the art accuracies. Output Video (Will be uploaded soon) |
