Generating TTS Based Adversarial Samples for Training Wake-Up Word Detection Systems Against Confusing Words.
Haoxu WangYan JiaZeqing ZhaoXuyang WangJunjie WangMing LiPublished in: Odyssey (2022)
Keyphrases
- word sense disambiguation
- n gram
- object detection
- training set
- training samples
- automatic transcription
- related words
- training examples
- supervised learning
- word recognition
- training corpus
- unknown words
- english words
- word meaning
- data sets
- multiword
- automatically generating
- text to speech
- lexical information
- word similarity
- prosodic features
- word frequencies