import musicntd.scripts.hide_code as hide

C:\Users\amarmore\AppData\Local\Continuum\anaconda3\envs\NTD_segmentation\lib\site-packages\librosa\util\decorators.py:9: NumbaDeprecationWarning: An import was requested from a module that has moved location.
Import requested from: 'numba.decorators', please update to use 'numba.core.decorators' or pin to Numba version 0.48.0. This alias will not be present in Numba version 0.50.0.
  from numba.decorators import jit as optional_jit
C:\Users\amarmore\AppData\Local\Continuum\anaconda3\envs\NTD_segmentation\lib\site-packages\librosa\util\decorators.py:9: NumbaDeprecationWarning: An import was requested from a module that has moved location.
Import of 'jit' requested from: 'numba.decorators', please update to use 'numba.core.decorators' or pin to Numba version 0.48.0. This alias will not be present in Numba version 0.50.0.
  from numba.decorators import jit as optional_jit

Distribution of segments' sizes in MIREX10¶

The study of the MIREX10 annotations in [1] shows that segments in this dataset are regular, and mostly centered around the size of 16 onbeats.

We replicated these results by studying the size in number of bars, more adapted to our context.

hide.repartition_bar_lengths_RWCPop()

Number of segments: 1619

We see in this histogram that most of the segment (more than a half) last 8 bars, and that numerous other segments last 4 bars. The other values are less represented. Hence, it should be interesting to enforce these sizes in our algorithm.

Thus, we modified our algorithm to include a regularization function, which favoures certain sizes of segments.

This regularization function only depends on the size of the segment, and is subtracted to the convolution cost. Hence, it's a penalty added to the raw convolution cost. Denoting $c_{b_1, b_2}$ the convolution cost as defined previously (see Notebook 1 or the Appendix which details our algorithm), the "regularized" cost is defined as $c'_{b_1,b_2} = c_{b_1,b_2} - \lambda p(b_2 - b_1 + 1) c_{k8}^{max}$, with:

$p(b_2 - b_1 + 1)$ the regularization function, which we will present in next part,
$\lambda$, a ponderation to handle the influence of this function,
$c_{k8}^{max}$, which represents the maximal convolution score on all intervals of size 8 of the current song, and its associated autosimilarity matrix. The idea of this score is to adapt the penalty to the specific shape of this autosimilarity. It can analogously be seen as a normalization of the raw convolution score: $\frac{c_{b_1,b_2}}{ c_{k8}^{max}} - \lambda p(b_2 - b_1 + 1)$

In this notebook, we will try to define the function $p$ and to study the influence of the parameter $\lambda$.

Definition of the regularization functions¶

We developped two types of regularization functions:

Symmetric functions centered on 8, as in [1]: the idea of this method is to favoure segments lasting 8 bars, as the majority of segments have this size, and to penalize all the other segments as the difference between their size and 8, raised to a certain power. Concretely, this results in:

$p(n) = |n - 8| ^{\alpha}$

with $n$ the size of the segment. Here, $\alpha$ is a parameter, and we will try
- $\alpha = \frac{1}{2}$
- $\alpha = 1$
- $\alpha = 2$
"Modulo functions": the idea of this method is to enforce specific sizes of segments, based on prior knowledge. They may be more empirical, but are also more adaptable to the need. Our main idea when developping this function was to favoure 8 wihout penalizing too much 4 or 16, that we know are current sizes (especially 4 in RWC Pop, as shown above). In addition, we considered that segments of even sizes should appear more often than segments of odd sizes in western pop music, which is less obvious in the distribution from above.

We will try 3 different types of function:
- "Favouring modulo 4": in this case:
  - if $n \equiv 0 \pmod 4, p(n) = 0$,
  - else, if $n \equiv 0 \pmod 2, p(n) = \frac{1}{2}$,
  - else, $p(n) = 1$.
- "Favouring 8, then modulo 4": in this case:
  - if $n = 8, p(n) = 0$,
  - else, if $n \equiv 0 \pmod 4, p(n) = \frac{1}{4}$,
  - else, if $n \equiv 0 \pmod 2, p(n) = \frac{1}{2}$,
  - else, $p(n) = 1$.
- "Favouring little segments of 8, then 4": in this case:
  - if $n > 12, p(n) = 100$ (we forbid the algorithm to select segments lasting more than 12 bars),
  - else, if $n = 8, p(n) = 0$,
  - else, if $n \equiv 0 \pmod 4, p(n) = \frac{1}{4}$,
  - else, if $n \equiv 0 \pmod 2, p(n) = \frac{1}{2}$,
  - else, $p(n) = 1$.

We will now test all these functions on the entire RWC Pop database.

First, we will fix the ranks to $T' = 32$ (rank for factor $H$) and $B' = 32$ (rank for factor $Q$), and search for the best $\lambda$ in this condition. In this condition, scores will only be plotted for the tolerance window of 0.5 seconds.
Secondly, we will keep the best parameter $\lambda$ for the ranks 32,32, and we will try several values of ranks with this $\lambda$ fixed. In this condition, we will show scores with both tolerances 0.5 seconds and 3 seconds, in that order.

As stated in the previous notebook, we now fix the subdivision to 96.

subdivision = 96

Estimating $\lambda$, with fixed ranks $T' = 32$ and $B' = 32$¶

Symmetric functions centered on 8¶

$\alpha = \frac{1}{2}$¶

param_range = [i/100 for i in range(0,100,5)]
hide.convolution_parameter_on_all_rwc(param_range, subdivision = subdivision, penalty_func = "sargentdemi")

c:\users\amarmore\desktop\projects\phd main projects\on git\code\tensor factorization\musicntd\autosimilarity_segmentation.py:43: RuntimeWarning: invalid value encountered in true_divide
  this_array = np.array([list(i/np.linalg.norm(i)) for i in this_array.T]).T

For $\alpha = \frac{1}{2}$, the best $\lambda$ (in our range) seems to be 0.4.

$\alpha = 1$¶

param_range = [i/1000 for i in range(0,100, 5)]
hide.convolution_parameter_on_all_rwc(param_range, subdivision = subdivision, penalty_func = "sargentun")

For $\alpha = 1$, the best $\lambda$ seems to be 0.09.

$\alpha = 2$¶

param_range = [i/1000 for i in range(0,20)]
hide.convolution_parameter_on_all_rwc(param_range, subdivision = subdivision, penalty_func = "sargentdeux")

For $\alpha = 2$, the best $\lambda$ seems to be 0.004.

Modulo functions¶

"Favouring modulo 4"¶

param_range = [i/10 for i in range(0,20)]
hide.convolution_parameter_on_all_rwc(param_range, subdivision = subdivision, penalty_func = "modulo4")

In this function, the best $\lambda$ seems to be 0.9.

"Favouring 8, then modulo 4"¶

param_range = [i/10 for i in range(0,20)]
hide.convolution_parameter_on_all_rwc(param_range, subdivision = subdivision, penalty_func = "modulo8")

In this function, the best $\lambda$ seems to be 1.

"Favouring little segments of 8, then 4"¶

param_range = [i/10 for i in range(0,20)]
hide.convolution_parameter_on_all_rwc(param_range, subdivision = subdivision, penalty_func = "moduloSmall8and4")

In this function, the best $\lambda$ seems to be 1.

Results with different ranks, $\lambda$ fixed¶

Now, we will fix $\lambda$ to the previous optimal value, and try different ranks for the decomposition.

For both $H$ and $Q$, we will try ranks in the range [12,16,20,24,28,32,36].

ranks_rhythm = [12,16,20,24,28,32,36]
ranks_pattern = [12,16,20,24,28,32,36]

Note: I conceed that tables are a bit long, you can jump directly to the conclusion to see what we considered to be the best outputs, and check if it's coherent in the tables afterwards.

Symmetric functions centered on 8¶

$\alpha = 1/2$¶

zero_five_chr, three_chr = hide.compute_ranks_RWC(ranks_rhythm,ranks_pattern,subdivision=subdivision, 
                                                  penalty_weight = 0.4, penalty_func = "sargentdemi")

$\alpha = 1$¶

zero_five_chr, three_chr = hide.compute_ranks_RWC(ranks_rhythm,ranks_pattern,subdivision=subdivision, 
                                                  penalty_weight = 0.09, penalty_func = "sargentun")

$\alpha = 2$¶

zero_five_chr, three_chr = hide.compute_ranks_RWC(ranks_rhythm,ranks_pattern,subdivision=subdivision, 
                                                  penalty_weight = 0.004, penalty_func = "sargentdeux")

Modulo functions¶

"Favouring modulo 4"¶

zero_five_chr, three_chr = hide.compute_ranks_RWC(ranks_rhythm,ranks_pattern, subdivision=subdivision,
                                                  penalty_weight = 0.9, penalty_func = "modulo4")

"Favouring 8, then modulo 4"¶

zero_five_chr, three_chr = hide.compute_ranks_RWC(ranks_rhythm,ranks_pattern, subdivision=subdivision,
                                                  penalty_weight = 1, penalty_func = "modulo8")

"Favouring little segments of 8, then 4"¶

zero_five_chr, three_chr = hide.compute_ranks_RWC(ranks_rhythm,ranks_pattern, subdivision=subdivision,
                                                  penalty_weight = 1, penalty_func = "moduloSmall8and4")

Conclusion¶

In conclusion, we systematically found that our best results at 0.5 seconds with the modulo functions are higher than the ones obtained with the symmetric functions centered on 8 (either when varying $\lambda$ or the ranks). At 3 seconds, it is less clear whether there is a best function, as best results are generally close.

In that sense, we chose to keep only the modulo functions, and more particularily "favouring 8, then modulo 4", as it gave the best results in this category (especially at 0.5 seconds).

References¶

[1] Sargent, G., Bimbot, F., & Vincent, E. (2016). Estimating the structural segmentation of popular music pieces under regularity constraints. IEEE/ACM Transactions on Audio, Speech, and Language Processing, 25(2), 344-358.

	Vrai Positifs	Faux Positifs	Faux Négatifs	Precision	Rappel	F mesure
0.0	10.9400	11.7800	7.8700	0.4876	0.5922	0.5259
0.05	10.4800	9.8400	8.3300	0.5229	0.5663	0.5347
0.1	10.2200	8.6800	8.5900	0.5459	0.5514	0.5413
0.15	10.0100	7.8700	8.8000	0.5641	0.5403	0.5450
0.2	9.8500	7.2200	8.9600	0.5799	0.5307	0.5479
0.25	9.7200	6.7200	9.0900	0.5917	0.5231	0.5495
0.3	9.5900	6.3800	9.2200	0.5999	0.5159	0.5495
0.35	9.5300	6.3200	9.2800	0.6016	0.5146	0.5494
0.4	9.4600	6.1300	9.3500	0.6069	0.5115	0.5502
0.45	9.3900	6.2300	9.4200	0.6028	0.5076	0.5463
0.5	9.2300	6.3200	9.5800	0.5950	0.4975	0.5368
0.55	9.1300	6.5200	9.6800	0.5854	0.4924	0.5298
0.6	9.2200	6.5500	9.5900	0.5864	0.4965	0.5330
0.65	9.2200	6.7400	9.5900	0.5800	0.4958	0.5295
0.7	9.2100	6.8600	9.6000	0.5753	0.4940	0.5268
0.75	9.1000	7.0000	9.7100	0.5664	0.4894	0.5206
0.8	9.0400	7.1700	9.7700	0.5616	0.4863	0.5167
0.85	8.9900	7.3500	9.8200	0.5563	0.4841	0.5131
0.9	9.0000	7.3500	9.8100	0.5554	0.4852	0.5133
0.95	8.9600	7.4600	9.8500	0.5518	0.4828	0.5103

	Vrai Positifs	Faux Positifs	Faux Négatifs	Precision	Rappel	F mesure
0.0	10.9400	11.7800	7.8700	0.4876	0.5922	0.5259
0.005	10.8000	10.9700	8.0100	0.5013	0.5840	0.5304
0.01	10.6900	10.5700	8.1200	0.5085	0.5783	0.5321
0.015	10.5900	10.2600	8.2200	0.5146	0.5724	0.5329
0.02	10.5500	9.9900	8.2600	0.5191	0.5700	0.5346
0.025	10.4800	9.7600	8.3300	0.5238	0.5657	0.5356
0.03	10.4800	9.5400	8.3300	0.5295	0.5658	0.5390
0.035	10.3100	9.3000	8.5000	0.5319	0.5568	0.5365
0.04	10.1800	9.1400	8.6300	0.5325	0.5498	0.5337
0.045	10.0900	8.9900	8.7200	0.5341	0.5443	0.5318
0.05	10.1000	8.7600	8.7100	0.5404	0.5444	0.5352
0.055	9.9600	8.6800	8.8500	0.5388	0.5373	0.5313
0.06	9.9200	8.5500	8.8900	0.5419	0.5359	0.5323
0.065	9.8700	8.3300	8.9400	0.5464	0.5330	0.5331
0.07	9.8400	8.2300	8.9700	0.5489	0.5309	0.5337
0.075	9.8500	8.0800	8.9600	0.5538	0.5310	0.5363
0.08	9.8000	7.9100	9.0100	0.5571	0.5283	0.5366
0.085	9.7900	7.7900	9.0200	0.5611	0.5276	0.5382
0.09	9.7700	7.6700	9.0400	0.5647	0.5266	0.5394
0.095	9.7000	7.6500	9.1100	0.5621	0.5226	0.5364

	Vrai Positifs	Faux Positifs	Faux Négatifs	Precision	Rappel	F mesure
0.0	10.9400	11.7800	7.8700	0.4876	0.5922	0.5259
0.001	10.8100	10.9800	8.0000	0.5023	0.5841	0.5312
0.002	10.6700	10.6800	8.1400	0.5059	0.5769	0.5305
0.003	10.6000	10.4600	8.2100	0.5094	0.5729	0.5310
0.004	10.5800	10.3000	8.2300	0.5129	0.5712	0.5327
0.005	10.4600	10.0500	8.3500	0.5161	0.5647	0.5321
0.006	10.3100	9.8600	8.5000	0.5173	0.5571	0.5299
0.007	10.2600	9.6600	8.5500	0.5206	0.5540	0.5308
0.008	10.1500	9.6600	8.6600	0.5181	0.5475	0.5268
0.009	10.0600	9.5600	8.7500	0.5185	0.5419	0.5246
0.01	9.9600	9.4500	8.8500	0.5187	0.5372	0.5222
0.011	9.8800	9.3800	8.9300	0.5182	0.5330	0.5201
0.012	9.8000	9.2200	9.0100	0.5199	0.5286	0.5192
0.013	9.7100	9.1100	9.1000	0.5196	0.5245	0.5172
0.014	9.6300	9.0400	9.1800	0.5203	0.5199	0.5155
0.015	9.5900	9.0200	9.2200	0.5196	0.5183	0.5145
0.016	9.5100	8.9600	9.3000	0.5192	0.5145	0.5125
0.017	9.4700	8.9200	9.3400	0.5199	0.5122	0.5117
0.018	9.4600	8.8500	9.3500	0.5214	0.5113	0.5119
0.019	9.4100	8.8400	9.4000	0.5203	0.5087	0.5100

	Vrai Positifs	Faux Positifs	Faux Négatifs	Precision	Rappel	F mesure
0.0	10.9400	11.7800	7.8700	0.4876	0.5922	0.5259
0.1	10.8700	10.2700	7.9400	0.5193	0.5877	0.5424
0.2	10.7200	9.2700	8.0900	0.5415	0.5794	0.5510
0.3	10.7600	8.3500	8.0500	0.5682	0.5802	0.5655
0.4	10.6300	7.7700	8.1800	0.5812	0.5737	0.5692
0.5	10.4000	7.6400	8.4100	0.5782	0.5607	0.5618
0.6	10.3800	7.2400	8.4300	0.5897	0.5589	0.5671
0.7	10.2200	7.2000	8.5900	0.5874	0.5499	0.5612
0.8	10.2600	6.9700	8.5500	0.5958	0.5527	0.5665
0.9	10.2700	6.7600	8.5400	0.6029	0.5534	0.5703
1.0	10.1000	6.7300	8.7100	0.5989	0.5455	0.5646
1.1	9.9200	6.7600	8.8900	0.5948	0.5357	0.5579
1.2	9.8500	6.6300	8.9600	0.5963	0.5316	0.5563
1.3	9.8200	6.5400	8.9900	0.6002	0.5310	0.5576
1.4	9.6200	6.6600	9.1900	0.5882	0.5204	0.5466
1.5	9.5700	6.5800	9.2400	0.5893	0.5183	0.5460
1.6	9.5500	6.5500	9.2600	0.5892	0.5170	0.5453
1.7	9.2500	6.7600	9.5600	0.5742	0.5030	0.5309
1.8	9.1500	6.8000	9.6600	0.5698	0.4965	0.5254
1.9	9.1600	6.7900	9.6500	0.5707	0.4970	0.5262

	Vrai Positifs	Faux Positifs	Faux Négatifs	Precision	Rappel	F mesure
0.0	10.9400	11.7800	7.8700	0.4876	0.5922	0.5259
0.1	10.8100	10.1700	8.0000	0.5207	0.5846	0.5418
0.2	10.6500	9.2300	8.1600	0.5424	0.5757	0.5490
0.3	10.6900	8.3300	8.1200	0.5680	0.5771	0.5635
0.4	10.4400	7.7600	8.3700	0.5774	0.5627	0.5620
0.5	10.3100	7.4400	8.5000	0.5837	0.5540	0.5611
0.6	10.2200	7.0400	8.5900	0.5942	0.5483	0.5634
0.7	10.0600	6.9400	8.7500	0.5925	0.5414	0.5591
0.8	10.0800	6.6800	8.7300	0.6023	0.5428	0.5644
0.9	10.0800	6.3200	8.7300	0.6146	0.5420	0.5698
1.0	10.1100	6.0400	8.7000	0.6266	0.5450	0.5767
1.1	9.9900	5.9800	8.8200	0.6252	0.5392	0.5728
1.2	9.8100	5.8800	9.0000	0.6242	0.5301	0.5677
1.3	9.7200	5.8800	9.0900	0.6213	0.5242	0.5631
1.4	9.6100	5.7800	9.2000	0.6202	0.5188	0.5591
1.5	9.3400	5.8500	9.4700	0.6118	0.5050	0.5474
1.6	9.2500	5.7000	9.5600	0.6148	0.4999	0.5459
1.7	9.0900	5.7800	9.7200	0.6058	0.4914	0.5372
1.8	8.9000	5.8800	9.9100	0.5974	0.4813	0.5277
1.9	8.6200	6.0600	10.1900	0.5851	0.4680	0.5145

	Résultats à 0.5 secondes	Vrai Positifs	Faux Positifs	Faux Négatifs	Precision	Rappel	F mesure
Rang Q:12	Rang H:12	7.8800	5.7800	10.9300	0.5773	0.4246	0.4825
	Rang H:16	7.8200	6.0000	10.9900	0.5634	0.4228	0.4766
	Rang H:20	8.0400	5.9000	10.7700	0.5755	0.4338	0.4886
	Rang H:24	8.1600	5.8700	10.6500	0.5798	0.4384	0.4922
	Rang H:28	8.2900	5.9900	10.5200	0.5837	0.4467	0.4991
	Rang H:32	7.9700	5.9300	10.8400	0.5715	0.4291	0.4837
	Rang H:36	8.2100	5.9900	10.6000	0.5766	0.4420	0.4938
Rang Q:16	Rang H:12	8.4100	5.9400	10.4000	0.5813	0.4531	0.5032
	Rang H:16	8.7000	5.8500	10.1100	0.5969	0.4688	0.5193
	Rang H:20	8.9600	5.7100	9.8500	0.6046	0.4825	0.5313
	Rang H:24	8.7900	5.8400	10.0200	0.5992	0.4743	0.5234
	Rang H:28	8.7500	5.8900	10.0600	0.5940	0.4696	0.5192
	Rang H:32	8.8200	5.6400	9.9900	0.6047	0.4750	0.5268
	Rang H:36	8.8700	5.9000	9.9400	0.5977	0.4760	0.5242
Rang Q:20	Rang H:12	8.9800	5.6000	9.8300	0.6159	0.4829	0.5354
	Rang H:16	9.0300	5.9800	9.7800	0.5980	0.4861	0.5302
	Rang H:20	9.0700	6.0000	9.7400	0.5995	0.4893	0.5336
	Rang H:24	9.1800	5.8600	9.6300	0.6077	0.4947	0.5393
	Rang H:28	9.1500	5.8900	9.6600	0.6063	0.4927	0.5373
	Rang H:32	9.1700	5.9000	9.6400	0.6078	0.4935	0.5384
	Rang H:36	9.1000	6.0200	9.7100	0.6025	0.4906	0.5349
Rang Q:24	Rang H:12	8.9100	6.0500	9.9000	0.5928	0.4813	0.5254
	Rang H:16	9.3000	5.9200	9.5100	0.6115	0.5025	0.5459
	Rang H:20	9.3400	6.0000	9.4700	0.6108	0.5045	0.5475
	Rang H:24	9.5000	6.0000	9.3100	0.6127	0.5135	0.5530
	Rang H:28	9.2800	6.0800	9.5300	0.6031	0.4994	0.5410
	Rang H:32	9.2900	5.9700	9.5200	0.6088	0.4999	0.5441
	Rang H:36	9.4000	6.0200	9.4100	0.6056	0.5060	0.5464
Rang Q:28	Rang H:12	9.1900	5.7500	9.6200	0.6141	0.4971	0.5440
	Rang H:16	9.1500	6.2500	9.6600	0.5900	0.4913	0.5310
	Rang H:20	9.3400	6.1100	9.4700	0.6018	0.5015	0.5421
	Rang H:24	9.4300	6.0400	9.3800	0.6066	0.5062	0.5466
	Rang H:28	9.5300	5.9500	9.2800	0.6135	0.5125	0.5534
	Rang H:32	9.3900	5.9000	9.4200	0.6129	0.5050	0.5484
	Rang H:36	9.6200	5.9300	9.1900	0.6169	0.5175	0.5578
Rang Q:32	Rang H:12	9.4000	5.7200	9.4100	0.6226	0.5063	0.5532
	Rang H:16	9.5000	5.8700	9.3100	0.6190	0.5123	0.5548
	Rang H:20	9.5200	6.0900	9.2900	0.6085	0.5107	0.5495
	Rang H:24	9.5600	6.1900	9.2500	0.6059	0.5140	0.5508
	Rang H:28	9.5900	6.1600	9.2200	0.6084	0.5151	0.5518
	Rang H:32	9.4600	6.1300	9.3500	0.6069	0.5115	0.5502
	Rang H:36	9.3400	6.3000	9.4700	0.5966	0.5028	0.5408
Rang Q:36	Rang H:12	9.1900	6.2600	9.6200	0.5944	0.4931	0.5338
	Rang H:16	9.2000	6.3300	9.6100	0.5909	0.4934	0.5325
	Rang H:20	9.3500	6.3600	9.4600	0.5948	0.5011	0.5392
	Rang H:24	9.4000	6.3600	9.4100	0.5963	0.5060	0.5425
	Rang H:28	9.3300	6.7600	9.4800	0.5808	0.5015	0.5334
	Rang H:32	9.3600	6.6700	9.4500	0.5815	0.5007	0.5333
	Rang H:36	9.3300	6.6200	9.4800	0.5845	0.5010	0.5354

	Résultats à 0.5 secondes	Vrai Positifs	Faux Positifs	Faux Négatifs	Precision	Rappel	F mesure
Rang Q:12	Rang H:12	7.9600	7.0900	10.8500	0.5374	0.4299	0.4705
	Rang H:16	7.9400	7.4400	10.8700	0.5234	0.4286	0.4643
	Rang H:20	8.0500	7.3700	10.7600	0.5309	0.4345	0.4711
	Rang H:24	8.2700	7.1600	10.5400	0.5419	0.4447	0.4817
	Rang H:28	8.1200	7.4100	10.6900	0.5324	0.4366	0.4733
	Rang H:32	7.9600	7.4200	10.8500	0.5251	0.4281	0.4644
	Rang H:36	8.1700	7.5500	10.6400	0.5283	0.4392	0.4733
Rang Q:16	Rang H:12	8.5400	7.2800	10.2700	0.5480	0.4608	0.4933
	Rang H:16	8.6400	7.2700	10.1700	0.5506	0.4650	0.4970
	Rang H:20	8.8900	7.2400	9.9200	0.5556	0.4799	0.5086
	Rang H:24	8.8500	7.2900	9.9600	0.5552	0.4774	0.5071
	Rang H:28	8.8000	7.3600	10.0100	0.5529	0.4752	0.5038
	Rang H:32	8.8600	7.0700	9.9500	0.5624	0.4793	0.5115
	Rang H:36	8.8500	7.2400	9.9600	0.5572	0.4753	0.5066
Rang Q:20	Rang H:12	9.0200	7.1900	9.7900	0.5669	0.4867	0.5171
	Rang H:16	9.0400	7.4200	9.7700	0.5578	0.4883	0.5140
	Rang H:20	9.1300	7.3900	9.6800	0.5595	0.4913	0.5169
	Rang H:24	9.2800	7.4000	9.5300	0.5645	0.5004	0.5238
	Rang H:28	9.1500	7.3300	9.6600	0.5639	0.4948	0.5201
	Rang H:32	9.0900	7.4600	9.7200	0.5583	0.4904	0.5152
	Rang H:36	9.1100	7.4400	9.7000	0.5593	0.4941	0.5177
Rang Q:24	Rang H:12	9.1600	7.2500	9.6500	0.5679	0.4947	0.5214
	Rang H:16	9.1200	7.3300	9.6900	0.5632	0.4930	0.5196
	Rang H:20	9.3300	7.5700	9.4800	0.5614	0.5037	0.5249
	Rang H:24	9.5900	7.4300	9.2200	0.5716	0.5185	0.5371
	Rang H:28	9.5900	7.4200	9.2200	0.5695	0.5185	0.5365
	Rang H:32	9.6000	7.2600	9.2100	0.5776	0.5175	0.5404
	Rang H:36	9.4700	7.5600	9.3400	0.5621	0.5087	0.5282
Rang Q:28	Rang H:12	9.2400	7.5300	9.5700	0.5580	0.5004	0.5207
	Rang H:16	9.2800	7.9800	9.5300	0.5462	0.4994	0.5159
	Rang H:20	9.3800	7.9600	9.4300	0.5481	0.5053	0.5198
	Rang H:24	9.4200	7.7600	9.3900	0.5552	0.5064	0.5239
	Rang H:28	9.5700	7.6000	9.2400	0.5617	0.5149	0.5315
	Rang H:32	9.4900	7.7400	9.3200	0.5533	0.5104	0.5257
	Rang H:36	9.5400	7.5700	9.2700	0.5631	0.5150	0.5322
Rang Q:32	Rang H:12	9.4800	7.5800	9.3300	0.5634	0.5124	0.5302
	Rang H:16	9.7400	7.3800	9.0700	0.5779	0.5228	0.5428
	Rang H:20	9.7800	7.5800	9.0300	0.5720	0.5247	0.5409
	Rang H:24	9.6900	7.8600	9.1200	0.5573	0.5215	0.5331
	Rang H:28	9.9500	7.6300	8.8600	0.5732	0.5349	0.5470
	Rang H:32	9.7700	7.6700	9.0400	0.5647	0.5266	0.5394
	Rang H:36	9.6700	7.9300	9.1400	0.5565	0.5217	0.5324
Rang Q:36	Rang H:12	9.4500	7.6300	9.3600	0.5605	0.5083	0.5278
	Rang H:16	9.5000	7.8100	9.3100	0.5543	0.5092	0.5253
	Rang H:20	9.6500	7.9500	9.1600	0.5569	0.5194	0.5317
	Rang H:24	9.7400	8.1000	9.0700	0.5530	0.5242	0.5325
	Rang H:28	9.6300	8.1900	9.1800	0.5477	0.5182	0.5266
	Rang H:32	9.6900	8.4300	9.1200	0.5392	0.5206	0.5243
	Rang H:36	9.6200	8.4900	9.1900	0.5376	0.5181	0.5222

	Résultats à 3 secondes	Vrai Positifs	Faux Positifs	Faux Négatifs	Precision	Rappel	F mesure
Rang Q:12	Rang H:12	10.6100	4.4400	8.2000	0.7192	0.5704	0.6270
	Rang H:16	10.7900	4.5900	8.0200	0.7147	0.5803	0.6316
	Rang H:20	10.9700	4.4500	7.8400	0.7276	0.5911	0.6433
	Rang H:24	11.0900	4.3400	7.7200	0.7336	0.5957	0.6486
	Rang H:28	11.0900	4.4400	7.7200	0.7321	0.5959	0.6486
	Rang H:32	10.6600	4.7200	8.1500	0.7067	0.5721	0.6231
	Rang H:36	11.1100	4.6100	7.7000	0.7224	0.5965	0.6448
Rang Q:16	Rang H:12	11.0900	4.7300	7.7200	0.7149	0.5982	0.6418
	Rang H:16	11.4800	4.4300	7.3300	0.7364	0.6183	0.6628
	Rang H:20	11.8000	4.3300	7.0100	0.7438	0.6362	0.6768
	Rang H:24	11.8400	4.3000	6.9700	0.7448	0.6376	0.6784
	Rang H:28	11.6500	4.5100	7.1600	0.7364	0.6282	0.6682
	Rang H:32	11.4500	4.4800	7.3600	0.7290	0.6165	0.6603
	Rang H:36	11.6700	4.4200	7.1400	0.7379	0.6266	0.6691
Rang Q:20	Rang H:12	11.6100	4.6000	7.2000	0.7321	0.6257	0.6661
	Rang H:16	11.9400	4.5200	6.8700	0.7394	0.6440	0.6795
	Rang H:20	12.0200	4.5000	6.7900	0.7396	0.6461	0.6813
	Rang H:24	12.0400	4.6400	6.7700	0.7357	0.6473	0.6799
	Rang H:28	12.1300	4.3500	6.6800	0.7514	0.6550	0.6907
	Rang H:32	11.7200	4.8300	7.0900	0.7194	0.6304	0.6631
	Rang H:36	11.9600	4.5900	6.8500	0.7369	0.6447	0.6788
Rang Q:24	Rang H:12	11.7300	4.6800	7.0800	0.7316	0.6324	0.6690
	Rang H:16	12.0100	4.4400	6.8000	0.7446	0.6474	0.6845
	Rang H:20	12.2000	4.7000	6.6100	0.7340	0.6548	0.6843
	Rang H:24	12.4400	4.5800	6.3700	0.7441	0.6678	0.6952
	Rang H:28	12.3500	4.6600	6.4600	0.7359	0.6639	0.6901
	Rang H:32	12.1800	4.6800	6.6300	0.7331	0.6554	0.6850
	Rang H:36	12.3000	4.7300	6.5100	0.7346	0.6598	0.6874
Rang Q:28	Rang H:12	12.0400	4.7300	6.7700	0.7282	0.6492	0.6776
	Rang H:16	12.3000	4.9600	6.5100	0.7266	0.6614	0.6845
	Rang H:20	12.5800	4.7600	6.2300	0.7363	0.6738	0.6957
	Rang H:24	12.4000	4.7800	6.4100	0.7315	0.6658	0.6892
	Rang H:28	12.4500	4.7200	6.3600	0.7348	0.6670	0.6914
	Rang H:32	12.3100	4.9200	6.5000	0.7223	0.6614	0.6832
	Rang H:36	12.5600	4.5500	6.2500	0.7420	0.6750	0.6989
Rang Q:32	Rang H:12	12.1400	4.9200	6.6700	0.7218	0.6528	0.6768
	Rang H:16	12.3900	4.7300	6.4200	0.7352	0.6654	0.6903
	Rang H:20	12.3900	4.9700	6.4200	0.7246	0.6648	0.6855
	Rang H:24	12.4900	5.0600	6.3200	0.7208	0.6706	0.6872
	Rang H:28	12.5300	5.0500	6.2800	0.7232	0.6727	0.6888
	Rang H:32	12.7000	4.7400	6.1100	0.7356	0.6820	0.7004
	Rang H:36	12.7100	4.8900	6.1000	0.7331	0.6822	0.6986
Rang Q:36	Rang H:12	12.3000	4.7800	6.5100	0.7286	0.6605	0.6857
	Rang H:16	12.6000	4.7100	6.2100	0.7352	0.6748	0.6961
	Rang H:20	12.7100	4.8900	6.1000	0.7334	0.6804	0.6980
	Rang H:24	12.6700	5.1700	6.1400	0.7207	0.6792	0.6917
	Rang H:28	12.7600	5.0600	6.0500	0.7245	0.6833	0.6953
	Rang H:32	12.9800	5.1400	5.8300	0.7246	0.6944	0.7016
	Rang H:36	12.8500	5.2600	5.9600	0.7199	0.6886	0.6964

	Résultats à 0.5 secondes	Vrai Positifs	Faux Positifs	Faux Négatifs	Precision	Rappel	F mesure
Rang Q:12	Rang H:12	8.2500	8.6300	10.5600	0.5001	0.4460	0.4629
	Rang H:16	8.3800	8.8500	10.4300	0.4963	0.4532	0.4652
	Rang H:20	8.4400	8.5200	10.3700	0.5102	0.4569	0.4732
	Rang H:24	8.4500	8.6000	10.3600	0.5100	0.4563	0.4732
	Rang H:28	8.4700	8.7900	10.3400	0.5041	0.4564	0.4702
	Rang H:32	8.4500	8.8100	10.3600	0.5010	0.4549	0.4680
	Rang H:36	8.6100	8.7100	10.2000	0.5108	0.4652	0.4786
Rang Q:16	Rang H:12	8.8400	9.0600	9.9700	0.5026	0.4766	0.4800
	Rang H:16	9.0500	9.0900	9.7600	0.5113	0.4883	0.4911
	Rang H:20	9.3900	9.1200	9.4200	0.5178	0.5061	0.5035
	Rang H:24	9.2500	9.1500	9.5600	0.5129	0.5005	0.4984
	Rang H:28	9.2100	9.1400	9.6000	0.5155	0.4976	0.4973
	Rang H:32	9.2000	9.2500	9.6100	0.5087	0.4957	0.4940
	Rang H:36	9.2000	9.0700	9.6100	0.5140	0.4950	0.4967
Rang Q:20	Rang H:12	9.5100	9.1200	9.3000	0.5222	0.5129	0.5085
	Rang H:16	9.6000	9.2900	9.2100	0.5163	0.5166	0.5081
	Rang H:20	9.4900	9.5000	9.3200	0.5106	0.5121	0.5030
	Rang H:24	9.6500	9.2500	9.1600	0.5193	0.5204	0.5109
	Rang H:28	9.7300	9.3700	9.0800	0.5201	0.5252	0.5138
	Rang H:32	9.8400	9.4400	8.9700	0.5203	0.5304	0.5175
	Rang H:36	9.7100	9.2000	9.1000	0.5204	0.5264	0.5154
Rang Q:24	Rang H:12	9.6800	9.4000	9.1300	0.5154	0.5233	0.5109
	Rang H:16	9.8100	9.4300	9.0000	0.5215	0.5304	0.5172
	Rang H:20	10.1400	9.8300	8.6700	0.5189	0.5486	0.5244
	Rang H:24	10.0400	9.8400	8.7700	0.5160	0.5441	0.5214
	Rang H:28	10.2200	9.8200	8.5900	0.5200	0.5529	0.5273
	Rang H:32	10.1200	10.0800	8.6900	0.5108	0.5453	0.5197
	Rang H:36	10.0400	10.0900	8.7700	0.5067	0.5421	0.5158
Rang Q:28	Rang H:12	9.9100	9.6000	8.9000	0.5177	0.5368	0.5178
	Rang H:16	10.0200	10.2900	8.7900	0.5018	0.5398	0.5118
	Rang H:20	10.0300	10.0400	8.7800	0.5065	0.5405	0.5151
	Rang H:24	10.1100	10.0500	8.7000	0.5094	0.5448	0.5191
	Rang H:28	10.3100	9.8800	8.5000	0.5196	0.5538	0.5281
	Rang H:32	10.4400	10.1300	8.3700	0.5123	0.5639	0.5299
	Rang H:36	10.3700	10.1400	8.4400	0.5151	0.5605	0.5283
Rang Q:32	Rang H:12	10.1100	9.6000	8.7000	0.5230	0.5474	0.5264
	Rang H:16	10.3700	9.9200	8.4400	0.5177	0.5571	0.5296
	Rang H:20	10.6400	9.9200	8.1700	0.5247	0.5728	0.5402
	Rang H:24	10.6900	10.4700	8.1200	0.5120	0.5744	0.5335
	Rang H:28	10.5600	10.3800	8.2500	0.5104	0.5680	0.5305
	Rang H:32	10.5800	10.3000	8.2300	0.5129	0.5712	0.5327
	Rang H:36	10.5300	10.5800	8.2800	0.5078	0.5692	0.5281
Rang Q:36	Rang H:12	10.1100	10.1600	8.7000	0.5062	0.5442	0.5176
	Rang H:16	10.4100	10.2600	8.4000	0.5104	0.5594	0.5269
	Rang H:20	10.5100	10.4400	8.3000	0.5093	0.5652	0.5287
	Rang H:24	10.4400	10.4800	8.3700	0.5035	0.5623	0.5249
	Rang H:28	10.4400	10.6200	8.3700	0.5006	0.5613	0.5227
	Rang H:32	10.5200	11.0100	8.2900	0.4957	0.5647	0.5215
	Rang H:36	10.3900	10.8900	8.4200	0.4942	0.5601	0.5189

	Résultats à 3 secondes	Vrai Positifs	Faux Positifs	Faux Négatifs	Precision	Rappel	F mesure
Rang Q:12	Rang H:12	11.1500	5.7300	7.6600	0.6786	0.5998	0.6254
	Rang H:16	11.4100	5.8200	7.4000	0.6819	0.6149	0.6355
	Rang H:20	11.4200	5.5400	7.3900	0.6949	0.6161	0.6413
	Rang H:24	11.5900	5.4600	7.2200	0.7018	0.6237	0.6491
	Rang H:28	11.5100	5.7500	7.3000	0.6891	0.6200	0.6409
	Rang H:32	11.3300	5.9300	7.4800	0.6770	0.6094	0.6296
	Rang H:36	11.5900	5.7300	7.2200	0.6894	0.6234	0.6437
Rang Q:16	Rang H:12	11.4800	6.4200	7.3300	0.6600	0.6193	0.6270
	Rang H:16	11.9000	6.2400	6.9100	0.6766	0.6418	0.6476
	Rang H:20	12.3200	6.1900	6.4900	0.6864	0.6642	0.6638
	Rang H:24	12.3000	6.1000	6.5100	0.6866	0.6630	0.6637
	Rang H:28	12.1500	6.2000	6.6600	0.6839	0.6558	0.6577
	Rang H:32	12.0900	6.3600	6.7200	0.6698	0.6500	0.6492
	Rang H:36	12.2000	6.0700	6.6100	0.6875	0.6555	0.6607
Rang Q:20	Rang H:12	12.1800	6.4500	6.6300	0.6730	0.6571	0.6536
	Rang H:16	12.4900	6.4000	6.3200	0.6770	0.6713	0.6634
	Rang H:20	12.3900	6.6000	6.4200	0.6693	0.6655	0.6564
	Rang H:24	12.3900	6.5100	6.4200	0.6731	0.6673	0.6586
	Rang H:28	12.6600	6.4400	6.1500	0.6835	0.6825	0.6715
	Rang H:32	12.4600	6.8200	6.3500	0.6631	0.6706	0.6566
	Rang H:36	12.4900	6.4200	6.3200	0.6732	0.6721	0.6623
Rang Q:24	Rang H:12	12.3800	6.7000	6.4300	0.6649	0.6679	0.6553
	Rang H:16	12.7000	6.5400	6.1100	0.6797	0.6843	0.6709
	Rang H:20	12.8700	7.1000	5.9400	0.6619	0.6925	0.6655
	Rang H:24	12.9900	6.8900	5.8200	0.6702	0.6999	0.6735
	Rang H:28	12.9300	7.1100	5.8800	0.6630	0.6964	0.6678
	Rang H:32	12.8700	7.3300	5.9400	0.6538	0.6928	0.6624
	Rang H:36	12.9500	7.1800	5.8600	0.6610	0.6961	0.6673
Rang Q:28	Rang H:12	12.6500	6.8600	6.1600	0.6657	0.6823	0.6622
	Rang H:16	12.8700	7.4400	5.9400	0.6511	0.6913	0.6595
	Rang H:20	13.0300	7.0400	5.7800	0.6633	0.6981	0.6701
	Rang H:24	13.0200	7.1400	5.7900	0.6607	0.7000	0.6700
	Rang H:28	13.0900	7.1000	5.7200	0.6649	0.7013	0.6722
	Rang H:32	13.0400	7.5300	5.7700	0.6457	0.7018	0.6636
	Rang H:36	13.1600	7.3500	5.6500	0.6586	0.7083	0.6717
Rang Q:32	Rang H:12	12.6000	7.1100	6.2100	0.6531	0.6787	0.6547
	Rang H:16	12.9500	7.3400	5.8600	0.6517	0.6963	0.6639
	Rang H:20	13.0700	7.4900	5.7400	0.6484	0.7014	0.6643
	Rang H:24	13.3000	7.8600	5.5100	0.6442	0.7138	0.6666
	Rang H:28	13.1600	7.7800	5.6500	0.6405	0.7066	0.6628
	Rang H:32	13.2400	7.6400	5.5700	0.6463	0.7111	0.6670
	Rang H:36	13.4000	7.7100	5.4100	0.6509	0.7200	0.6726
Rang Q:36	Rang H:12	12.9900	7.2800	5.8200	0.6538	0.6986	0.6662
	Rang H:16	13.2600	7.4100	5.5500	0.6525	0.7107	0.6715
	Rang H:20	13.4800	7.4700	5.3300	0.6571	0.7217	0.6783
	Rang H:24	13.2400	7.6800	5.5700	0.6432	0.7108	0.6669
	Rang H:28	13.5500	7.5100	5.2600	0.6535	0.7259	0.6793
	Rang H:32	13.5400	7.9900	5.2700	0.6411	0.7253	0.6720
	Rang H:36	13.4900	7.7900	5.3200	0.6444	0.7239	0.6735

	Résultats à 3 secondes	Vrai Positifs	Faux Positifs	Faux Négatifs	Precision	Rappel	F mesure
Rang Q:12	Rang H:12	10.0500	3.6100	8.7600	0.7384	0.5412	0.6161
	Rang H:16	10.3800	3.4400	8.4300	0.7510	0.5595	0.6327
	Rang H:20	10.5600	3.3800	8.2500	0.7608	0.5687	0.6428
	Rang H:24	10.5500	3.4800	8.2600	0.7548	0.5675	0.6388
	Rang H:28	10.7200	3.5600	8.0900	0.7574	0.5781	0.6469
	Rang H:32	10.1400	3.7600	8.6700	0.7288	0.5451	0.6156
	Rang H:36	10.6600	3.5400	8.1500	0.7522	0.5727	0.6418
Rang Q:16	Rang H:12	10.6600	3.6900	8.1500	0.7426	0.5733	0.6392
	Rang H:16	11.1500	3.4000	7.6600	0.7678	0.5996	0.6656
	Rang H:20	11.4500	3.2200	7.3600	0.7798	0.6172	0.6816
	Rang H:24	11.2700	3.3600	7.5400	0.7738	0.6067	0.6720
	Rang H:28	11.2800	3.3600	7.5300	0.7705	0.6064	0.6715
	Rang H:32	11.1400	3.3200	7.6700	0.7663	0.5984	0.6651
	Rang H:36	11.4300	3.3400	7.3800	0.7755	0.6129	0.6772
Rang Q:20	Rang H:12	11.1600	3.4200	7.6500	0.7671	0.5998	0.6655
	Rang H:16	11.6000	3.4100	7.2100	0.7728	0.6246	0.6827
	Rang H:20	11.6500	3.4200	7.1600	0.7753	0.6266	0.6861
	Rang H:24	11.5900	3.4500	7.2200	0.7735	0.6229	0.6820
	Rang H:28	11.7300	3.3100	7.0800	0.7822	0.6335	0.6916
	Rang H:32	11.3500	3.7200	7.4600	0.7546	0.6105	0.6670
	Rang H:36	11.5500	3.5700	7.2600	0.7679	0.6221	0.6795
Rang Q:24	Rang H:12	11.1900	3.7700	7.6200	0.7488	0.6028	0.6603
	Rang H:16	11.8000	3.4200	7.0100	0.7805	0.6374	0.6940
	Rang H:20	11.7200	3.6200	7.0900	0.7673	0.6315	0.6861
	Rang H:24	12.0200	3.4800	6.7900	0.7786	0.6477	0.6995
	Rang H:28	11.8300	3.5300	6.9800	0.7715	0.6366	0.6904
	Rang H:32	11.6800	3.5800	7.1300	0.7674	0.6282	0.6843
	Rang H:36	11.9300	3.4900	6.8800	0.7752	0.6422	0.6959
Rang Q:28	Rang H:12	11.4200	3.5200	7.3900	0.7658	0.6151	0.6753
	Rang H:16	11.7900	3.6100	7.0200	0.7668	0.6342	0.6870
	Rang H:20	11.9000	3.5500	6.9100	0.7727	0.6394	0.6927
	Rang H:24	12.0000	3.4700	6.8100	0.7742	0.6444	0.6962
	Rang H:28	12.0200	3.4600	6.7900	0.7748	0.6457	0.6973
	Rang H:32	11.7000	3.5900	7.1100	0.7653	0.6287	0.6831
	Rang H:36	12.0700	3.4800	6.7400	0.7762	0.6485	0.7000
Rang Q:32	Rang H:12	11.5700	3.5500	7.2400	0.7635	0.6223	0.6787
	Rang H:16	11.8700	3.5000	6.9400	0.7759	0.6390	0.6931
	Rang H:20	11.7100	3.9000	7.1000	0.7512	0.6281	0.6766
	Rang H:24	12.0100	3.7400	6.8000	0.7642	0.6457	0.6929
	Rang H:28	11.9500	3.8000	6.8600	0.7611	0.6413	0.6881

	Résultats à 3 secondes	Vrai Positifs	Faux Positifs	Faux Négatifs	Precision	Rappel	F mesure
Rang Q:12	Rang H:12	9.4300	3.2900	9.3800	0.7532	0.5095	0.5958
	Rang H:16	9.8800	3.3400	8.9300	0.7608	0.5353	0.6172
	Rang H:20	9.9300	3.3600	8.8800	0.7625	0.5379	0.6195
	Rang H:24	9.9400	3.3200	8.8700	0.7600	0.5369	0.6185
	Rang H:28	10.0100	3.3700	8.8000	0.7600	0.5407	0.6212
	Rang H:32	9.7100	3.5000	9.1000	0.7468	0.5240	0.6044
	Rang H:36	10.0200	3.2300	8.7900	0.7633	0.5405	0.6228
Rang Q:16	Rang H:12	10.3400	3.4700	8.4700	0.7599	0.5593	0.6321
	Rang H:16	10.8000	3.4600	8.0100	0.7712	0.5812	0.6522
	Rang H:20	11.0900	3.3800	7.7200	0.7796	0.5983	0.6657
	Rang H:24	10.9900	3.3500	7.8200	0.7787	0.5938	0.6641
	Rang H:28	11.0100	3.4000	7.8000	0.7746	0.5953	0.6626
	Rang H:32	10.8300	3.5100	7.9800	0.7676	0.5818	0.6515
	Rang H:36	11.0600	3.5700	7.7500	0.7671	0.5948	0.6606
Rang Q:20	Rang H:12	10.7100	3.6300	8.1000	0.7586	0.5761	0.6441
	Rang H:16	11.3400	3.4500	7.4700	0.7799	0.6119	0.6746
	Rang H:20	11.3900	3.8200	7.4200	0.7639	0.6125	0.6690
	Rang H:24	11.4300	3.6800	7.3800	0.7675	0.6153	0.6734
	Rang H:28	11.6200	3.6400	7.1900	0.7733	0.6265	0.6820
	Rang H:32	11.1400	3.9200	7.6700	0.7499	0.5980	0.6549
	Rang H:36	11.5200	3.6700	7.2900	0.7687	0.6207	0.6766
Rang Q:24	Rang H:12	11.0200	3.8100	7.7900	0.7540	0.5936	0.6536
	Rang H:16	11.3000	3.6500	7.5100	0.7642	0.6100	0.6690
	Rang H:20	11.5200	3.8500	7.2900	0.7584	0.6208	0.6732
	Rang H:24	11.7500	3.8900	7.0600	0.7641	0.6316	0.6813
	Rang H:28	11.8500	3.9200	6.9600	0.7591	0.6365	0.6826
	Rang H:32	11.6900	3.8900	7.1200	0.7590	0.6282	0.6785
	Rang H:36	11.7400	4.0200	7.0700	0.7579	0.6320	0.6799
Rang Q:28	Rang H:12	11.2500	3.8200	7.5600	0.7548	0.6060	0.6613
	Rang H:16	11.7400	4.0700	7.0700	0.7545	0.6324	0.6775
	Rang H:20	11.7600	3.9900	7.0500	0.7520	0.6320	0.6776
	Rang H:24	12.0100	3.9400	6.8000	0.7612	0.6449	0.6894
	Rang H:28	12.0000	4.0800	6.8100	0.7546	0.6448	0.6865
	Rang H:32	11.7100	4.2000	7.1000	0.7406	0.6288	0.6715
	Rang H:36	11.9000	4.0500	6.9100	0.7478	0.6416	0.6816
Rang Q:32	Rang H:12	11.4300	4.0000	7.3800	0.7439	0.6146	0.6625
	Rang H:16	11.8600	4.1800	6.9500	0.7465	0.6363	0.6780
	Rang H:20	11.7300	4.0700	7.0800	0.7454	0.6297	0.6740
	Rang H:24	11.9200	4.4300	6.8900	0.7357	0.6393	0.6746
	Rang H:28	11.8900	4.2600	6.9200	0.7429	0.6377	0.6771