Table 1. Summary of the seven...

Table 1.

Summary of the seven transcription conditions used for training, characterized by word error rate (WER), character error rate (CER) and sentence error rate (S.Err)

Transcriptions	SNR (dB)	$C E R ↓$	$W E R ↓$	S.Err $↓$
Name	Interval	(%)	(%)	(%)
Clean	–	0.0	0.0	0.0
Wer-070	$\infty$	3.3	7.0	59.6
Wer-162	[−10, 0)	9.9	16.2	81.4
Wer-242	[−15, 0)	15.7	24.2	86.8
Wer-358	[−20, 0)	24.8	35.8	90.1
Wer-437	[−20, −5)	30.9	43.7	94.2
Wer-548	[−20, −10)	39.4	54.8	97.8

Transcriptions	SNR (dB)	$C E R ↓$	$W E R ↓$	S.Err $↓$
Name	Interval	(%)	(%)	(%)
Clean	–	0.0	0.0	0.0
Wer-070	$\infty$	3.3	7.0	59.6
Wer-162	[−10, 0)	9.9	16.2	81.4
Wer-242	[−15, 0)	15.7	24.2	86.8
Wer-358	[−20, 0)	24.8	35.8	90.1
Wer-437	[−20, −5)	30.9	43.7	94.2
Wer-548	[−20, −10)	39.4	54.8	97.8

[ViewLarge]

Sharing Unavailable