HMM-TDNN + pNorm + speed up/down speech

Reported on 2 benchmarks across 1 task

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

Audio2 results

Speech Recognitiononswb_hub_500 WER fullSWBCH
Percentage error
19.3
best: 6.8 (IBM (LSTM+Conformer encoder-decoder))
Speech RecognitiononSwitchboard + Hub500
Percentage error
12.9
best: 4.3 (IBM (LSTM+Conformer encoder-decoder))