Lute Acoustic Quality Evaluation and Note Recognition Based on the Softmax Regression BP Neural Network

Table 1

Average accuracy of varying the number of training samples for different feature inputs.

Input	10 sets of samples	30 sets of samples	50 sets of samples	70 sets of samples	90 sets of samples	110 sets of samples
MFCC	91.63	95.09	95.48	95.64	94.92	95.70
CQT	92.57	97.33	96.81	97.81	98.41	98.50
MFCC + CQT	96.16	96.1	97.24	97.21	98.68	99.14
MFCC + CC	94.79	95.7	95.09	96.11	96.15	96.72
CQT + CC	92.16	97.25	97.77	98.06	98.55	98.82
MFCC + CQT + CC	94.61	94.78	97.89	98.7	99.36	99.68