Joint Time-Frequency Scattering for Audio Classification
arXiv:1512.02125 · doi:10.1109/MLSP.2015.7324385
Abstract
We introduce the joint time-frequency scattering transform, a time shift invariant descriptor of time-frequency structure for audio classification. It is obtained by applying a two-dimensional wavelet transform in time and log-frequency to a time-frequency wavelet scalogram. We show that this descriptor successfully characterizes complex time-frequency phenomena such as time-varying filters and frequency modulated excitations. State-of-the-art results are achieved for signal reconstruction and phone segment classification on the TIMIT dataset.
arXiv comment6 pages, 2 figures in IEEE 25th International Workshop on Machine Learning for Signal Processing (MLSP), 2015. Sept. 17-20. Boston, USA
References in corpus (2)
Cited by in corpus (12)
- Understanding Deep Convolutional Networks
- Robust sound event detection in bioacoustic sensor networks
- Wavelet Scattering on the Pitch Spiral
- Gabor frames and deep scattering networks in audio processing
- On Time-frequency Scattering and Computer Music
- The Shape of RemiXXXes to Come: Audio Texture Synthesis with Time-frequency Scattering
- Time-Frequency Localization Using Deep Convolutional Maxout Neural Network in Persian Speech Recognition
- Energy Propagation in Scattering Convolution Networks Can Be Arbitrarily Slow
- Deep Autoencoders: From Understanding to Generalization Guarantees
- Cortical Features for Defense Against Adversarial Audio Attacks
- Interferometric Graph Transform for Community Labeling
- Sparse Multi-Family Deep Scattering Network