Cleanformer: A microphone array configuration-invariant, streaming, multichannel neural enhancement frontend for ASR.
Joseph CaroselliArun NaranayanTom O'MalleyPublished in: CoRR (2022)
Keyphrases
- network architecture
- automatic speech recognition
- data streams
- real time
- neural network
- image processing
- image enhancement
- real time streaming
- video streaming
- streaming data
- back end
- affine invariant
- invariant properties
- single channel
- invariant features
- associative memory
- affine transformation
- multiscale
- multi channel
- hidden markov models
- noisy environments
- multiresolution
- hebbian learning
- neural fuzzy
- information retrieval
- speech retrieval