Federating Dynamic Models using Early-Exit Architectures for Automatic Speech Recognition on Heterogeneous Clients.
Mohamed Nabih AliAlessio BruttiDaniele FalavignaPublished in: CoRR (2024)
Keyphrases
- dynamic model
- automatic speech recognition
- speech recognition
- speech signal
- experimental data
- word error rate
- hidden markov models
- broadcast news
- conversational speech
- spoken words
- speech retrieval
- noisy environments
- spontaneous speech
- recognition errors
- speech corpus
- multiple models
- word recognition
- machine learning
- multiscale
- web services
- feature selection