Development and Evaluation of an Intelligent System for Calibrating Karaoke Lyrics Based on Fuzzy Petri Nets

Figures & data

Figure 1. Register change.

Figure 2. Dynamics change.

Figure 3. Articulation change.

Figure 4. Length change.

Figure 5. Timbre change.

Figure 6. HLFPN model of illustrative example.

Figure 7. All layers in IEEE 1599 (Wick, Hartelt, and Puppe 2020).

Figure 8. A phrase.

Figure 9. Block diagram of HLFPN model.

Figure 10. The type of membership degrees for input parameter.

Figure 11. The type of membership degrees for technical decision.

Table 1. Membership degrees of input parameter.

Input Parameter	Low		High
$f_{register}$	0.2	0.7	0.3	0.8
$f_{dynamics}$	0.2	0.7	0.3	0.8
$f_{length}$	0.2	0.7	0.3	0.8

Table 2. Membership degrees of technical decision.

Performance			Segmentation
0.0	0.2	0.4	0.3	0.5	0.7

Figure 12. The activity diagram of fuzzy reasoning.

Figure 13. The HLFPN model representing six fuzzy production rules.

Table 3. Description of parameters.

Name of Parameter	Description of Parameter
$f_{register}$	Denotes the value of $f_{register}$ in input place P1,
$f_{dynamics}$	Denotes the value of $f_{dynamics}$ in input place P2.
$f_{length}$	Denotes the value of $f_{length}$ in input place P3.
D	Denotes a decision in an output place P4.
H, L	Denote high and low fuzzy sets, respectively.
S, W	Denote strong and weak fuzzy sets, respectively.
$V(ti),i=1,2,3,\dots ,6$	Denotes the fuzzy relational matrix of $f_{register}$, $f_{dynamics}$, $f_{length}$ and decision.
H’, L’	Denote high and low fuzzy sets of input value.

Figure 14. Layers in IEEE 1599.

Figure 15. Description in general layer of IEEE 1599.

Figure 16. The related files in general layer of IEEE 1599.

Figure 17. Layers in IEEE 1599.

Figure 18. Deviation in automated system.

Figure 19. The actual score.

Table 4. Experimental results.

No.	Number of Phrases	Number of Phrases Correctly Detected	Number of Phrases Deviated	Number of Phrases Detected	Recall (%)	Precision (%)
1	43	39	3	42	90.70	92.86
2	46	39	4	43	84.78	90.70
3	49	45	3	48	91.84	93.75
4	48	42	3	45	87.50	93.33
5	22	19	2	21	86.36	90.48
6	34	31	2	33	91.18	93.94
7	27	25	2	27	92.59	92.59
8	34	31	3	34	91.18	91.18
9	49	45	3	48	91.84	93.75
10	44	40	3	43	90.91	93.02
11	24	22	1	23	91.67	95.65
12	40	35	3	38	87.50	92.11
13	39	36	2	38	92.31	94.74
14	27	24	2	26	88.89	92.31
15	45	39	3	42	86.67	92.86
16	43	40	2	42	93.02	95.24
17	48	43	3	46	89.58	93.48
18	36	33	2	35	91.67	94.29
19	39	35	3	38	89.74	92.11
20	32	29	2	31	90.63	93.55
21	48	44	2	46	91.67	95.65
22	44	39	4	43	88.64	90.70
23	42	38	3	41	90.48	92.68
24	43	39	3	42	90.70	92.86
25	53	47	5	52	88.68	90.38
26	36	33	2	35	91.67	94.29
27	49	45	3	48	91.84	93.75
28	31	28	2	30	90.32	93.33
29	35	32	2	34	91.43	94.12
30	53	47	4	51	88.68	92.16
31	60	54	4	58	90.00	93.10
32	44	40	3	43	90.91	93.02
33	36	32	3	35	88.89	91.43
34	59	53	4	57	89.83	92.98
35	45	41	3	44	91.11	93.18
36	38	34	3	37	89.47	91.89
37	40	36	3	39	90.00	92.31
38	42	38	3	41	90.48	92.68
39	46	42	3	45	91.30	93.33
40	50	45	4	49	90.00	91.84
41	61	56	4	60	91.80	93.33
42	46	42	4	46	91.30	91.30
43	47	43	3	46	91.49	93.48
44	22	20	2	22	90.91	90.91
45	40	36	3	39	90.00	92.31
46	37	33	3	36	89.19	91.67
47	32	32	0	32	100.0	100.0
48	36	33	2	35	91.67	94.29
49	58	52	4	56	89.66	92.86
50	34	30	2	32	88.24	93.75
Average					90.46	92.78

Table 5. Results of functional comparison.

Functions Approaches	Correct Phrase Detection	Deviated Phrase Detection	Missing Phrase Detection	Recall (%)	Precision (%)
Our Proposed	∨	∨	∨	90.46	92.78
X. Wu’s et al	∨	∨			∨
S. Bilbao’s et al	∨	∨			∨

Wick, C., A. Hartelt, and F. Puppe. 2020. Lyrics recognition and syllable assignment of medieval music manuscripts, Procs. of 2020 17th. International Conference on Frontiers in Handwriting Recognition (ICFHR), Dortmund, Germany, 1–6.

 Google Scholar