Analysis of Two-Piano Teaching Assistant Training Based on Neural Network Model Sound Sequence Recognition

Table 2

Comparison of AUC results under nonmatching noise.

Input noise type	−5 dB			0 dB			5 dB
	F2	Ba	Avg.	F2	Ba	Avg.	F2	Ba	Avg.
Sohn	82.6	76.2	78.5	90.4	82.7	85.1	91.8	89.3	90.3
LSTM-VAD	93.4	70.8	83.6	97.2	86.3	90.8	97.6	94.9	96.5
LSTM-MTL	95.9	78.3	86.4	98.3	87.5	93.2	98.5	97.2	97.6