Hybrid CNN-embedding fusion with MFCC-SVM for speech emotion recognition: Random vs actor-wise evaluation on CREMA-D
Parveen Kumari, Yogita Yashveer Raghav, Vimmi Kochher, Prashanth kumar katta, Abhilasha A, Prabhakar M, Jayanthiladevi A, Fikir Gizachew Belete · 2026
Read Hybrid CNN-embedding fusion with MFCC-SVM for speech emotion recognition: Random vs actor-wise evaluation on CREMA-D by Parveen Kumari, Yogita Yashveer Raghav, Vimmi Kochher, Prashanth kumar katta, Abhilasha A, Prabhakar M, Jayanthiladevi A, Fikir Gizachew Belete (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/plos-pone-0355238.txt.
Description
Parveen Kumari and a large team compared three ways of teaching a computer to hear emotion in speech. Each way was tested under two rules. Under the easy rule, the same actors appear in training and testing. Under the honest rule, test voices are strangers. Their 2026 study used a standard acted emotion corpus. It combined handcrafted sound features with deep network embeddings in a lightweight hybrid.
The hybrid won under both rules. Every method lost ground when voices changed, exposing how inflated easy-split scores are. The team's real contribution is discipline. It used identical preprocessing, repeated runs, and class-by-class error audits. A small web tool lets readers try the model live. Speech emotion systems headed for call centers and tutoring need exactly this kind of honest benchmarking.
Catalog record
| CREATOR | Parveen Kumari, Yogita Yashveer Raghav, Vimmi Kochher, Prashanth kumar katta, Abhilasha A, Prabhakar M, Jayanthiladevi A, Fikir Gizachew Belete |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 44855 |
| LICENSE | CC_BY |
| RIGHTS | CC BY |
| SOURCE | PLOS |