[KongchangAI]
Concept端到端声学路线

End-to-End Speech Processing

A speech processing paradigm that bypasses the traditional cascade of speech recognition followed by text processing; a single neural network maps directly from raw audio waveforms or acoustic features to the target task output, and in principle can exploit prosodic and articulatory information.

Timeline (last 90 days)

Nov 6

The end-to-end approach requires time-aligned annotations of audio and keywords, and such data are far scarcer than 'audio plus transcribed text', which makes large-scale deployment difficult.

Unverified50%

All Facts (1)

Source Articles