Table 1.
Summary of different methodologies used for SER
| No. | Dataset | Methodology | Results (accuracy) | Author |
|---|---|---|---|---|
| 1 | IVR customer care domain, database from WoZ data collectiona | SVM | 79%, 75% | Polzehl et al. (2011) |
| 2 | IEMOCAP corpusb | RNN | 63.5% | Mirsamadi et al. (2017) |
| 3 | EMO-DB, VAM, and TUM AVIC | SVM | 51.6% | Deng et al. (2013) |
| 4 | Berlin EmoDB and IEMOCAP | CNN, LSTM | 95.33%, 95.89% on Berlin EmoDB; 89.16%, 52.14% on IEMOCAP | Zhao et al. (2019) |
| 5 | EMO-DB | SVM | 74.4% | Deb and Dandapat (2016) |
| 6 | EMO-DB and IEMOCAP | Bidirectional LSTM and CNN | 82.35% | Pandey et al. (2019) |
| 7 | (UMSSEDc) and (RAVDESSd) | Four models for binary classification | 64.29% | Zhang et al. (2016) |
| 8 | RAVDESS | CNN | 66.41%. | Jannat et al. (2018) |
| 9 | RAVDESS | SVM, NN | 78.75%, 89.16% | Tomba et al. (2018) |
| 10 | RAVDESS | SVM | 75.69% | Bhavan et al. (2019) |
| 11 | GeWEC | Universum AE | 59.3% | Deng et al. (2018) |
| 12 | GeWEC | SSAE | 51.6% | Deng et al. (2017) |
| 13 | SAVEE | SVM, DSM, AE | 69.84%, 68.25%, 73.01% | Aouani and Ben Ayed (2018) |