The Kears implementation of "Cross-Cultural Music Emotion Recognition by Adversarial Discriminative Domain Adaptation".
Citation: Yi-Wei Chen, Yi-Hsuan Yang, and Homer H. Chen, "Cross-cultural music emotion recognition by adversarial discriminative domain adaptation," Proc. IEEE Int. Conf. Machine Learning and Applications (ICMLA), December 2018.
| Adaptation | Timbre | Pitch | Rhythm | Timbre + Pitch | Timbre + Rhythm | Rhythm + Pitch | Timbre + Pitch + Rhythm | |
|---|---|---|---|---|---|---|---|---|
| Valence | - | 0.03 | 0.08 | 0.04 | 0.08 | 0.05 | 0.08 | 0.09 |
| V | 0.21 | 0.18 | 0.06 | 0.22 | 0.22 | 0.17 | 0.23 | |
| Arousal | - | 0.72 | 0.69 | 0.39 | 0.74 | 0.68 | 0.67 | 0.74 |
| V | 0.73 | 0.65 | 0.28 | 0.76 | 0.65 | 0.49 | 0.71 |
- Tansfer audio files into 22.05KHz and clip into 29 seconds
Call function: Transfer_funcs.audio_to_wav()
- Generate combined audio files and labels for training
Call function: Transfer_funcs.wav_to_npy()
- Extract log-mel spectrogram [1]
Call function: model_structure.extract_melspec()
- Extract pitch salience representation [2]
Use predict_on_audio.py
ex: python predict_on_audio.py '/mnt/data/Wayne/Dataset/AMG_1608_wav@22050Hz' 'pitch' '/mnt/data/Wayne/AMG_1608_pitch+lw@22050Hz'
- Extract autocorrelation-based tempogram [3]
Use MATLAB-Tempogram-Toolbox_1.0/test_TempogramToolbox.m
- Transfer different features into the same npy format
Use Extract_features.py
- Within-dataset experiment
Use AMG1608_CV.py
- Cross-dataset experiment
Use AMG1608_train.py
- Cross-dataset experiment
Use WADDA_S_AMG1608_T_CH818.py
- Test for all the experiments in the paper and generate results in exel files
Use Multi_fusion_pred_S_AMG1608_T_CH818_find_by_loss.py
- K. Choi, D. Joo, and J. Kim, "Kapre: On-gpu audio preprocessing layers for a quick implementation of deep neural network models with keras," arXiv preprint arXiv:1706.05781, 2017.
- R. M. Bittner et al., “Deep salience representations for f0 estimation in polyphonic music,” in Proc. Int. Soc. Music Information Retrieval, pp. 23–27, 2017.
- P. Grosche, M. Muller, and F. Kurth, “Cyclic tempogram—a midlevel tempo representation for music signals,” in IEEE Trans. Acoustics Speech Signal Process., pp. 5522–5525, 2010.