Table 1 Comparison between SER baselines and proposed model
From: Speech emotion recognition based on Graph-LSTM neural network
Model | UA (%) | WA (%) | Condition |
|---|---|---|---|
DCNN 2020 [41] | - | 64.3 | 4490 utterances |
ResNet34 2021 [42] | 61.61 | 66.02 | |
ADNN + SVM 2019 [43] | - | 65.01 | |
Graph baselines | |||
PATCHY-SAN 2016 [11] | 56.27 | 60.34 | |
PATCHY-Diff 2018 [11] | 58.71 | 63.23 | |
Compact SER 2021 (cycle) [11] | 62.27 | 65.29 | |
Ours (Mean pooling) | 59.16 | 68.15 | |
Ours (Weighted pooling) | 65.39 | 71.83 | |
LSTM-GIN 2022 [46] | 65.53 | 64.65 | 5531 utterances |
CoGCN 2022 [33] | 63.67 | 62.64 | |
GA-GRU 2020 [25] | 63.8 | 62.27 | |
Ours (Mean pooling) | 68.65 | 68.11 | |