Spracherkennung
Definition
Spracherkennung (ASR) transkribiert Audio in Text. Related areas include speaker identification, speech synthesis (TTS), and spoken language understanding.
Es überbrückt multimodal (audio as one modality) and NLP (output is text). Modern ASR is mostly end-to-end neural; self-supervised pretraining (z. B. wav2vec 2.0) reduces the need for huge labeled datasets. Deployed in voice assistants, captions, and meeting tools.
Funktionsweise
Audio (waveform or mel spectrogram) is converted to features (z. B. filter banks, learned representations). An acoustic model (z. B. conformer, wav2vec 2.0 encoder) maps features to frame- or segment-level representations. A decoder (CTC, RNN-T, or attention-based) erzeugt text (characters or subwords). Modern systems werden oft end-to-end (waveform or features → text in one model). Self-supervised pretraining on unlabeled audio (z. B. wav2vec) then Feinabstimmung on labeled ASR data improves robustness and reduces labeled data needs.
Anwendungsfälle
Speech technologies gelten, wenn the input or output is audio: transcription, assistants, and speaker or synthesis systems.
- Automatic speech recognition (ASR) for transcription and captions
- Voice assistants and spoken dialogue systems
- Speaker identification and speech synthesis (TTS)