A Method to Reveal Speaker Identity in Distributed ASR Training, and How to Counter IT.
Trung DangOm ThakkarSwaroop RamaswamyRajiv MathewsPeter ChinFrançoise BeaufaysPublished in: ICASSP (2022)
Keyphrases
- experimental evaluation
- training samples
- objective function
- segmentation method
- speaker recognition
- feed forward neural networks
- training process
- speech recognition
- synthetic data
- detection method
- high accuracy
- significant improvement
- computational complexity
- probabilistic model
- computational cost
- pairwise
- preprocessing
- pattern recognition
- training phase
- multi agent
- similarity measure