SA-SOT: Speaker-Aware Serialized Output Training for Multi-Talker ASR.
Zhiyun FanLinhao DongJun ZhangLu LuZejun MaPublished in: CoRR (2024)
Keyphrases
- automatic speech recognition
- speech recognition
- simulated annealing
- training set
- training process
- discriminative training
- speaker recognition
- speech signal
- test set
- hybrid algorithm
- training algorithm
- online learning
- genetic algorithm ga
- training samples
- audio visual
- distributed systems
- speaker verification
- hidden markov models
- data sets