SpecMix : A Mixed Sample Data Augmentation Method for Training with Time-Frequency Domain Features.
Gwantae KimDavid K. HanHanseok KoPublished in: Interspeech (2021)
Keyphrases
- synthetic data
- input data
- training samples
- test data
- data sets
- feature set
- prior knowledge
- random samples
- classification accuracy
- high dimensional data
- missing data
- informative features
- sample set
- training examples
- pairwise
- feature selection
- data samples
- feature extraction
- clustering method
- model selection
- image data
- data sources
- feature space
- database
- linear svm
- receiver operating characteristic curves
- irrelevant features
- selecting relevant features
- training phase
- extracted features
- feature subset
- statistical methods
- classification method
- support vector machine svm
- data mining techniques
- wavelet transform
- image features
- training set
- data structure
- training data