HMM-TDNN + iVectors

Reported on 3 benchmarks across 1 task

Note: results are matched by exact model name. Different papers may use the same name for different model variants.

Audio3 results

Speech Recognitiononswb_hub_500 WER fullSWBCH
Percentage error
17.1
best: 6.8 (IBM (LSTM+Conformer encoder-decoder))
Speech RecognitiononLibriSpeech test-clean
Word Error Rate (WER)· uses extra data
4.8
best: 0.985 (United Med ASR)
Speech RecognitiononSwitchboard + Hub500
Percentage error
11
best: 4.3 (IBM (LSTM+Conformer encoder-decoder))