Summary of automatic discrimination of Music and vocals

As shown in table 1 and table 2 below, the vocal test set contains: TV shows, storytelling, storytelling, talk crosstalk, audio radio, etc. the Music test set contains: Chinese dance, R&b, western ballad, Chinese rock, rap, 1614 mixed album and so on.

Table 1 Test concentration, the average vocal recognition rate of 89.31%, the average recognition rate of music in 95.95%. Crosstalk and the results of the story is poor, because the test concentrated in the scene of the crosstalk is poor, noisy environment, spectrum confusion, and storytelling has a more obvious musical background accompaniment.

Table 1 music and vocals automatic classification results

	Correct number	Number	Accuracy rate
Television	59	66	0.893939
Story	69	88	0.784091
Storytelling	181	186	0.973118
Crosstalk	73	96	0.760417
Audio Radio	41	44	0.930233
1800 Sound	96	101	0.950495
Total	518	580	0.893103
1800 Music	1614	1690	0.95503
Chinese dance	54	54	1
R&b	44	44	1
Western Ballads	80	80	1
Chinese Rock	50	52	0.961538
Rap	104	108	0.962963
Total	1946	2028	0.959566
Reasonable threshold value	0.5

Table 22 Layer Filter Automatic classification results

	Correct number	Number	Accuracy rate
Television	62	66	0.939394
Story	72	88	0.818182
Storytelling	182	186	0.978495
Crosstalk	77	96	0.802083
Audio Radio	41	43	0.953488
1800 Sound	96	101	0.950495
Total	530	580	0.913793
1800 Music	1592	1688	0.943128
Chinese dance	54	54	1
R&b	43	44	0.977273
Western Ballads	80	80	1
Chinese Rock	49	52	0.942308
Rap	104	108	0.962963
Total	1922	2026	0.948667
Reasonable threshold value	0.5

Summary of automatic discrimination of Music and vocals