EEG – based study on sleep quality improvement by using music

 Al-176
pha, sometimes that value was higher than 50% at one177
epoch, the subject began to sleep, at stage N1.178
K-means algorithm would randomly classify training 179
data into two clusters, respectively creating two cen- 180
troids of each cluster. After the data have been clus- 181
tered, it is necessary to find a boundary line between 182
the two clusters. This line is based on the definition 183
of K-means, which is a perpendicular bisector of two 184
centroids. Perpendicular bisector equation: 185
y=
x2 x1
y2 y1  (xmidX)+midY (2)
Whereas: 186
A(x1;y1) and B(x2;y2) were the centroid of “Wake” 187
and “Sleep” cluster, respectively. 188
M = (midX ; midY ): midpoint of A and B. 189
Set the variable test: 190
test =
x2 x1
y2 y1  (xmidX)+midY  y
(3)
Using (3) and assign (x,y) to test. If test<0, that point 191
belonged to cluster “Wake” andwas assigned the value 192
“1”. And vice versa, if test>0, that point belonged to 193
cluster “Sleep” and the value was “0”. As shown in Fig- 194
ure 2, the blue line was the perpendicular bisector that 195
divided training data into two clusters. A green point 196
stood for one epoch in the analysis data. 197
The matrix result “0” and “1” showed the transition 198
between “1” and “0” is the epoch number. The SL 199
value was computed by applying the following for- 200
mula: 201
t =
epoch number30
60
min (4)
RESULTS ANDDISCUSSION 202
Filtered data 203
There were totally of 14 samples, including 7 samples 204
with no intervention and 7 samples with music inter- 205
vention. The sampling frequency of the collected data 206
was 2000 Hz. 207
It can be seen from Figure 3 that the data were re- 208
moved the 50 Hz noise and fixed baseline drift prob- 209
lem by Wavelet transform. The signal was filtered in 210
the range from 0.5 Hz to 35 Hz, and converted from 211
time domain to frequency domain by Fourier trans- 212
form. Mother Wavelet Transform was “db7” because 213
its waveform matches the waveform of the occipital 214
region19. So, the O2 channel was used to investigate 215
the result. 216
3
Un
co
rre
cti
on
 pr
oo
f
Science & Technology Development Journal – Engineering and Technology, 3(SI):xxx-xxx
Figure 2: Boundary line and analysis data
Figure 3: De-noised signal by usingWavelet Transform
4
Un
co
rre
cti
on
 pr
oo
f
Science & Technology Development Journal – Engineering and Technology, 3(SI):xxx-xxx
Sleep latency (SL)217
Sleep onset is defined as the start of the first epoch218
when the subject starts falling asleep. Sleep latency219
is the time lights out to the first epoch of any sleep220
measured in min15. The shorter the sleep latency, the221
better the sleep quality. And the aim of this study is to222
use music to shorten the sleep latency value.223
In this study, we decided to focus on Theta wave224
and Alpha wave, which characterize for sleep study.225
Alpha wave characterizes human physiological alert-226
ness, while Theta wave appears when people feel227
sleepy. If there is a state change between these two228
waves (power ratio of Alpha decreases and power ra-229
tio ofTheta increases), the subject goes to sleep - sleep230
onset. From the moment of turning off the light and231
starting to measure until the subject falls asleep, that232
is the value of SL, in minute.233
It can be also seen from Figure 4 each symbol in was234
stood for one epoch which was 30-second long. The235
power ratio of Alpha and Theta were marked ‘*’ and236
‘O’, respectively. The horizontal axis was time axis237
(minute) and the vertical one was the power ratio R.238
Thepower ratio of Alpha reached over 0.8 at about the239
first 20minutes while the power ratio ofThetawas just240
around 0.1 at the same time. The intersection of two241
waves was the time subject fell to sleep. The other in-242
tersection was not investigated because the study had243
just focused on the fall to sleep time of the subject. The244
SL at the experiment without music intervention was245
estimated in range 21 - 23 minutes.246
Figure 5 shows the sleep late cy at the musical exper-247
iment on the same subject. The SL value was guessed248
in range 13.5 – 14 minutes.249
Classification250
The above results were evaluated by visualization and251
estimation so that it was necessary to develop amodel252
of predictive result, using the K-means clustering al-253
gorithm. The classification was classified based on the254
flowchart in Figure 1.255
Table 1 shows the accuracy of the K-means clustering256
algorithm. The accuracy was quite high (the lowest257
was 93.75%) and three data sets with 100% accuracy258
were three training data sets. This study investigated259
in naptime, therefore the subjects woke up randomly.260
So, the length of each experiment was different, which261
caused incorrect accuracy.262
Statistical analysis263
To test for differences between with and without mu-264
sic intervention, paired - samples for means t-test265
model was used. Statistical analysis impact of music266
to shorten the value of SL.267
Table 1: Accuracy of K-Means clustering algorithm
Sample Total of epoch Accuracy (%)
K_01_0204 121 100
K_02_2503 63 100
K_03_0304 81 93.83
K_04_2803 50 96.00
K_05_2603 102 99.17
K_06_2504 35 97.14
K_07_1804 32 93.75
C_01_0404 121 100
C_02_2703 60 98.41
C_03_0504 120 95.83
C_04_2903 126 95.24
C_05_2703 120 97.06
C_06_2804 100 94.00
C_07_1904 90 95.55
Figure 6 shows the sleep latency value of each subject 268
in two experiments: with andwithoutmusic interven- 269
tion. There are many different sleep latency values be- 270
cause each SL value depends on each person. To illus- 271
trate this, subject 1 reported in the survey that he did 272
not have a habit of napping, and music helped him 273
to fall asleep easier. Besides, subject 3 usually had a 274
break in the afternoon, for this reason, SL value was 275
not much different between the two experiments. In 276
sum, this research measures the change of sleep la- 277
tency by using their favorite music and shows the re- 278
sult that music can improve sleep in people who can- 279
not sleep in naptime. 280
Using ANOVA, the result: 281
• SL in normal experiment: mean = 9.0714 min, t 282
Stat = 3.0039, P (one-tail) = 1.9432, SD = 6.0853. 283
• SL in musical experiment: mean = 5.6423 min, t 284
Stat = 3.0039, P (one-tail) = 1.9432, SD = 3.4405. 285
It can clearly be seen that music helps to shorten SL 286
value of each subject, which makes people fall asleep 287
easier. But the SD value was high due to some reasons 288
below: 289
• The number of samples was limited. 290
• The SL value difference between subjects was 291
highly depended on the habit of having nap 292
sleep. 293
• The laboratory environment is not suitable for 294
some subjects owing to the enclosed space of 295
Faraday cage and noise. 296
5
Un
co
rr
cti
on
 pr
oo
f
Science & Technology Development Journal – Engineering and Technology, 3(SI):xxx-xxx
Figure 4: Sleep latency at the experiment without music intervention
Figure 5: Sleep latency at the experiment with music intervention
User interface297
To develop user interface by using GUIDE in MAT-298
LAB that helps the data were analyzed more conve-299
nient. Figure 7 shows the GUIDE developed inMAT-300
LAB with some main button to analyze data:301
- CHOOSE FILE: choose one data file in (*.txt) format302
- Options: Fs (Hz), Wavelet Name, From Detail – To303
Detail (analytical level based on NYQUIST rule)304
- Epoch (s): epoch duration and the program will cal-305
culate the total of epoch in that data file.306
- Epoch number: select the epoch to analysis and plot 307
the result. 308
- Slider: slip to the epoch number instead of typing 309
the epoch number. 310
- PLOT: the data were filtered and analyzed by using 311
Options. Then analyzed data were plotted on three 312
graphs to evaluate the results. 313
- Sleep latency: plot the power ratio of Theta and Al- 314
pha and export a dialog box of SL value as shown in 315
Figure 8. 316
6
Un
co
rre
cti
on
 pr
oo
f
Science & Technology Development Journal – Engineering and Technology, 3(SI):xxx-xxx
Figure 6: The sleep latency value of each subject in two experiments: with and without music intervention.
Figure 7: User interface GUIDE
7
Un
co
rre
cti
on
 pr
oo
f
Science & Technology Development Journal – Engineering and Technology, 3(SI):xxx-xxx
Figure 8: Sleep latency in GUIDE
CONCLUSION317
The present study had built an interface user to ana-318
lyze EEG signal using FFT, Wavelet, power ratio and319
K-means clustering algorithm. Instead of measur-320
ing SL by visualization or available self-reported ques-321
tionnaire, this study is empirical research that com-322
bines both the above methods. A new point of the323
research is Machine learning in analyzing data to cal-324
culate the SL value.325
The main aim of this study was to confirm that mu-326
sic affected to shorten the SL value – one factor in327
improving sleep quality, which helped the subject fall328
asleep easier and faster by using favoritemusic of each329
subject.330
However, there were some shortcomings that needed331
to be improved. The number of subjects and experi-332
ments was limited so that the result was not accurate.333
In the near future, this research can bemeasuredmore334
samples and limited the noise. Besides investigating335
each person’s response to certain types of music.336
ACKNOWLEDGMENT337
We acknowledge the support of time and facilities338
from Ho Chi Minh City University of Technology339
(HCMUT), VNU–HCM for this study.340
LIST OF ABBREVIATION341
AASM: American Academy of Sleep Medicine.342
EEG: Electroencephalogram.343
EMG: electromyogram.344
EOG: electrooculogram. 345
ESS: Epworth Sleepiness Scale. 346
FFT: Fast Fourier Transform. 347
MAR: Music – Assisted Relaxation. 348
PAI: Personality Assessment Inventory. 349
PSQI: Pittsburgh Sleep Quality Index. 350
SL: Sleep Latency. 351
AUTHOR S’ CONTRIBUTIONS 352
All authors contributed equally to this work. All au- 353
thors have read and agreed to the published version of 354
the manuscript. 355
CONFLICT OF INTEREST 356
We declare that there is no conflict of whatsoever in- 357
volved in publishing this research. 358
REFERENCES 359
1. Rosekind MR, et al. Alertness management: strategic naps 360
in operational settings. European Sleep Research Society. 361
1995;(4):62–66. PMID: 10607214. Available from: https://doi. 362
org/10.1111/j.1365-2869.1995.tb00229.x. 363
2. Chiung-Yu H, En-Ting C, Hui-Ling L. Use of integrative 364
medicine approaches for treating adults with sleep distur- 365
bances. Applied Nursing Research. 2018;43:49–55. PMID: 366
30220363. Available from: https://doi.org/10.1016/j.apnr.2018. 367
06.016. 368
3. Kira VJ, Julian K, Poul J, Peter V. Music for insomnia in 369
adults (Review). John Wiley & Sons, Ltd, Jylland. 2015;PMID: 370
26270746. Available from: https://doi.org/10.1002/14651858. 371
CD010459.pub2. 372
4. Gerrit N, Bea T, Bert L, Giel H. Music-assisted relaxation to im- 373
prove sleep quality: meta-analysis. Journal of AdvancedNurs- 374
ing. 2009;(65):1356–1364. PMID: 19456998. Available from: 375
https://doi.org/10.1111/j.1365-2648.2009.04982.x. 376
8
Un
co
rre
cti
n p
roo
f
Science & Technology Development Journal – Engineering and Technology, 3(SI):xxx-xxx
5. Tan LP. The Effects of Background Music on Quality of Sleep377
in Elementary School Children. Journal of Music Therapy.378
2004;XLI(2):128–150. PMID: 15307813. Available from: https:379
//doi.org/10.1093/jmt/41.2.128.380
6. La’szlo’ H, Johanna T, Ro’bert B. Music improves sleep qual-381
ity in students. Journal of Advanced Nursing. 2008;(62):327–382
335. PMID: 18426457. Available from: https://doi.org/10.1111/383
j.1365-2648.2008.04602.x.384
7. Fan F, Yingshi Z, Jun H, Jiayi C, Qiyu J, Xiaojuan L, Qingchun Z,385
Bo-an L. Can Music Improve Sleep Quality in Adults with Pri-386
mary Insomnia? A Systematic Review and Network Primary387
Insomnia? A Systematic Review and Network Meta-analysis.388
International Journal of Nursing Studies. 2017;p. 15–20.389
8. Nantawachara J, Yodchanan W. A Novel Insight of Effects of a390
3-Hz Binaural Beat on Sleep Stages during Sleep. Frontiers in391
Human Neuroscience. 2018;p. 7–10. PMID: 30319382. Avail-392
able from: https://doi.org/10.3389/fnhum.2018.00387.393
9. Hui-Ling L, Marion G. Music improves sleep quality in older394
adults. Journal of Advanced Nursing. 2004;(49):241. PMID:395
15660547. Available from: https://doi.org/10.1111/j.1365-396
2648.2004.03281.x.397
10. Stanley EL, Robert DO. Lack of efficacy of music to improve398
sleep: A polysomnographic and quantitive EEG analysis. In-399
ternational Journal of Psychophysiology. 2007;(63):232–239.400
PMID: 17123654. Available from: https://doi.org/10.1016/j.401
ijpsycho.2006.10.004.402
11. Abhijeet DD, Avani AS, Seethalakshmi R, Ajita SN. Effect of403
Indian classical music on quality of sleep in depressed pa-404
tients: A randomized controlled trial. Nordic Journal of Music405
Therapy. 2014;18(1):70–78. Available from: https://doi.org/10.406
1080/08098130802697269.407
12. Weina Z, Junjun Z, Xiaojun D, Changle Z, Yuanye M, Dan X.408
Crossmodal effects of Guqin and piano music on selective409
attention: An event-related potential study. Neuroscience410
Letters. 2009;(446):21–26. PMID: 19766172. Available from:411
https://doi.org/10.1016/j.neulet.2009.09.026.412
13. Santostasi G. EEG Analysis on Brain.fm (Sleep). [Online] [Ac-413
cessed 1 December 2018]. 2015;Available from: https://brain.414
fm.415
14. Daniel JB, Charles FR, Timothy HM, Susan RB, and David416
JK. The Pittsburgh Sleep Quality Index: A New Instrument417
for Psychiatric Practice and Research. Psychiatry Research.418
1988;(28):193–213. Available from: https://doi.org/10.1016/419
0165-1781(89)90047-4.420
15. Conrad I, Ancoli-Israel S, Andrew LC Jr, Stuart FQ. The ASSM421
Manual for the Scoring of Sleep and Associated Events.422
Westchester: American Academy of Sleep Medicine. 2007;.423
16. NicoletOnevEEGSystem.NatusNeurology Incorporated,Mid-424
dleton, Wisconsin. 2014;.425
17. Marion G, Bradford LP, Safaa GS, Chi-Chun C, SandraFP, Defo-426
ria L. Cultural Differences inMusic Chosen for Pain Relief. Jour-427
nal of Holistic Nursing. 2000;18(3):245–260. PMID: 11847812.428
Available from: https://doi.org/10.1177/089801010001800306.429
18. Hanan AA, Faris AJ. Optimal Mother Wavelet Function for430
EEG Signal Analyze Based on Packet Wavelet Transform.431
International Journal of Scientific & Engineering Research.432
2017;8(2):1222–1227.433
19. John LS, BenjaminG. Biosignal andMedical Image Processing.434
Boca Raton: CRC Press. 2014;p. 540–548.435
9
Un
co
rr
cti
on
 pr
oo
f