ArticlePDF Available

Abstract and Figures

Obstructive sleep apnea (OSA) is the most common form of different types of sleep-related breathing disorders. It is characterized by repetitive cessations of respiratory flow during sleep, which occurs due to a collapse of the upper respiratory airway. OSA is majorly undiagnosed due to the inconvenient Polysomnography (PSG) testing procedure at sleep labs. This paper introduces an automated approach towards identifying the presence of sleep apnea based on the acoustic signal of respiration. The characterization of breathing sound was carried by Voice Activity Detection (VAD) algorithm, which is used to measure the energy of the acoustic respiratory signal during breath and breath hold. The performance of our classification algorithm is tested on real respiratory signals and the experimental results show that the VAD is useful as a predictive tool for the segmentation of breath into sound and silence segments. Moreover, the system we developed can be used as a basis for future development of a tool for OSA screening.
Content may be subject to copyright.
Apnea Detection Based on
Respiratory Signal Classification
Laiali Almazaydeh, Khaled Elleithy, Miad Faezipour and Ahmad Abushakra
Department of Computer Science and Engineering
University of Bridgeport
Bridgeport, CT 06604, USA
{lalmazay, elleithy, mfaezipo, aabushak}@bridgeport.edu
Abstract Obstructive sleep apnea (OSA) is the most common
form of different types of sleep-related breathing disorders. It is
characterized by repetitive cessations of respiratory flow during
sleep, which occurs due to a collapse of the upper respiratory
airway. OSA is majorly undiagnosed due to the inconvenient
Polysomnography (PSG) testing procedure at sleep labs. This
paper introduces an automated approach towards identifying the
presence of sleep apnea based on the acoustic signal of
respiration. The characterization of breathing sound was carried
by Voice Activity Detection (VAD) algorithm, which is used to
measure the energy of the acoustic respiratory signal during
breath and breath hold. The performance of our classification
algorithm is tested on real respiratory signals and the
experimental results show that the VAD is useful as a predictive
tool for the segmentation of breath into sound and silence
segments. Moreover, the system we developed can be used as a
basis for future development of a tool for OSA screening.
Keywords: sleep apnea, OSA, PSG, VAD, respiration signal.
I. INTRODUCTION
Sleep apnea (SA) in the form of Obstructive sleep apnea
(OSA) is becoming the most common respiratory disorder
during sleep, which is characterized by cessations of airflow to
the lungs. These cessations in breathing must last more than 10
seconds to be considered an apnea event. Apnea events may
occur 5 to 30 times an hour and may occur up to four hundred
times per night in those with severe SA [1].
The most frequent night symptoms of SA can include
snoring, nocturnal arousals, sweating and restless sleep.
Moreover, like all sleeping disorders, symptoms of sleep apnea
do not occur just during the night. Daytime symptoms can
range from morning headaches, depression, impaired
concentration and excessive sleepiness which cause mortality
from traffic and industrial accidents. However, these symptoms
are not definitive to detect SA syndrome [2] [3].
In fact, SA is not a problem to be taken lightly, since it is
associated with a major risk factor of health implications and
increased cardiovascular disease and sudden death. It has been
linked to irritability, depression, sexual dysfunction, high blood
pressure (hypertension), learning and memory difficulties, in
addition to stroke and heart attack [2] [3]. Several treatment
options for OSA patients include weight loss, positional
therapy, oral appliances, surgical procedures and continuous
positive airway pressure (CPAP). CPAP is a common and
effective treatment especially for patients with moderate to
severe OSA. CPAP devices are masks worn during sleep that
improves oxygen saturation and reduces sleep fragmentation
[4].
Statistics show that around 100 million people worldwide,
and in the US from 18 to 50 million people, are suspected to
have OSA. This is while more than 80% of which remain
undiagnosed [5]. The trouble of having examinations
discourages patients prone to OSA undergo at the overnight
clinical research through polysomnographic data.
Polysomnography (PSG) is a complicated procedure and
certain way of assessing the OSA problem. Complete PSG
includes the monitoring of the breath airflow, respiratory
movement, oxygen saturation (SpO
2
), body position,
electroencephalography (EEG), electromyography (EMG),
electrooculography (EOG), and electrocardiography (ECG) [6].
However, PSG has received many criticisms from some
researchers. This is due to several reasons, including first, the
inconvenience since it requires the patient to be connected to
numerous sensors and to stay in hospital for one night. Second,
it is expensive. The average cost for a PSG is $2,625 due to the
need for the study to take place in a specially equipped medical
facility, in addition to the requirement of having a sleep lab
staff overnight, trained in ‘scoring’ the resultant measurements
manually. Third, a long wait list of up to 6 months is caused by
limited availability of PSG [7].
According to the American Academy of Sleep Medicine
(AASM), the Apnea-Hypopnea Index (AHI) is used to describe
the number of complete and partial apnea events per hour of
sleep and it is calculated to assess OSA syndrome severity.
OSA severity is usually determined as follows: AHI 5-15
indicates mild, 15-30 indicates moderate and over 30 indicates
severe OSA syndrome. Therefore, patients are diagnosed with
OSA if they have five or more apnea events per hour of sleep
during a full night sleep period [8].
However, new simplified methods for diagnosis and
screening of OSA are needed, in order to have a major benefit
of the treatment on OSA outcomes. In this work, we develop an
efficient algorithm for automatic classification of respiratory
signal to detect abnormalities in breathing or breathing
cessations. We use Voice Activity Detection (VAD) to classify
respiratory signals into normal respiration and sleep apnea. The
detection process would check for apnea attack for a time
period of fifteen seconds or more.
In the following sections, we glance at a variety of sleep
apnea detection methods. Section III, gives a general
description of VAD. In section IV, the methodology of our
proposed system is described. Section V demonstrates the
results of our system. Then, we conclude our paper in section
VI, and highlight some directions for future research.
II. RELATED WORK
Over the past few years most of the related research has
focused on presenting methods for the automatic processing of
different statistical features of different signals such as thorax
and abdomen effort signals, nasal air flow, oxygen saturation,
electrical activity of the heart (ECG), and electrical activity of
the brain (EEG) for the detection of SA.
In our previous published research, we developed a Neural
Network (NN) as a predictive tool for OSA using SpO2 signal
and evaluated its effectiveness [9]. In addition, in [10] we
further developed a model based on a linear kernel Support
Vector Machines (SVMs) using a selective set of RR-interval
features from short duration epochs of the ECG signal. The
results show that our automated classification system can
recognize epochs of SA with a high degree of accuracy,
approximately 96.5%
As OSA is generally caused by a blocked of the airflow
airway and it is characterized by repetitive episodes of
breathing cessation, the respiratory signal recording analysis
during sleep becomes very valuable in order to estimate
respiratory flow and distinguish the changes in the breathing
pattern of the patient. Then, to provide additional and
complementary information, other biological signal data
measurements such as ECG and SpO2 could be bridged to
analyze sleep data, as clinical experience indicates that an
apneic event is frequently accompanied by a fall in the blood
oxygen saturation (SpO2) [11], and cyclic variations in the
duration of a heartbeat (ECG); this consists of bradycardia
during apnea followed by tachycardia upon its cessation [12].
Recently, based on the SpO2 and tracheal breathing sound
recording analysis during sleep, the study in [13] reports a new
fully automatic technology for OSA detection. Different
parameters were investigated to distinguish the breathing level
during each individual apnea event. Therefore, in the first step,
the drops (more than 4%) and rises of the SpO2 signal were
marked, then, the total energy of the tracheal sound segments
within the periods between a drop and the following rise in the
SpO2 were found. After collection of data, each parameter was
then fuzzified with a sigmoid function and the fuzzy outputs
were added together to classify the sound signals. The results
show high sensitivity and specificity values of more than 90%
in differentiating normal respiration from disordered breathing
in patients.
Several studies for the non-invasive sensing of the
respiration rate are based either on the measurement of the
nasal airflow or on the evaluation of the respiration movements
(abdominal or chest respiration signals). Using mean absolute
amplitude analysis of the combination of the thoracic and the
abdominal signals in [14] showed that both signals are able to
indicate the occurrence of SA events. The analysis
identification results achieved a sensitivity ranged from 70.29-
86.25% and the specificity values ranged from 74.82 to
90.09%.
It has been reported that snoring is a common finding in
people with OSA. OSA is generally caused by a blocked of the
airflow airway. Therefore, the snoring must be due to the
vibration of soft tissues when the airflow stimulates the ill
structure in the upper airway during sleep [15]. Of all methods
for diagnosing OSA, the formants estimation method is most
widely used. The formants information contains the essential
acoustic properties of the upper airway. It has been discovered
by studies that there is a correlation between the state of the
upper airway and the first formant frequency. A narrower upper
airway is usually led to a higher first formant frequency.
Andrew et al. [16] and [17] proposed fixed formant frequency
thresholds to detect the hypopneic snores which must be higher
than that of the typical ones.
Various portable monitor devices already exist in the
market. SleepStrip
TM
is one of the carriages available in home
sleep test diagnostic devices. This device has to be worn for a
minimum of five hours of sleep, and the actual device is placed
on the individual’s face where the two flow sensors (oral and
nasal thermistors) are placed just below the nose and above the
upper lip to capture the breath of the patient. For all samples
combined, sensitivity and specificity values ranged from 80-
86% and 57-86% respectively [18].
Even though most of the related studies yielded promising
initial results, more improvement is needed, as it either requires
physical attachment to a user or may be unreliable. Therefore,
since apnea is a condition where patient pauses breathing; this
can be of great concern for detecting breathing through the
sound signal where an appropriate alarm can be released upon
its cessation.
III. VOICE ACTIVITY DETECTION THE PRINCIPLE
In this work, the principle of voice activity detection
(VAD) algorithm employed to detect the presence or absence
of apnea on real breathing signals is described.
Voice Activity Detector plays an important role in speech
processing techniques such as speech coding [19], speech
enhancement, and speech recognition [20]. Other examples
include cellular radio systems (GSM and CDMA based) [21],
hands-free telephony [22], VoIP applications and echo
cancellation.
VAD relies on measurement of features from speech which
yield highly in differentiating between voiced and unvoiced
segments, where the regions of voice information within a
given audio signal are referred to as ‘voice-active’ segments
and the pauses between talking are called ‘silence’ or ‘voice-
inactive’ segments. Therefore, the performance trade-offs of
VAD algorithm are made by maximizing the detection rate of
active speech while minimizing the false detection rate of
inactive segments [23].
The most important part in VAD classifier is feature
extraction, from which different regions in the audio signal
can be separated. Common features used in the VAD detection
process are cepstral coefficient [24], spectral entropy [19],
zero-crossing rate [20, 25], least square periodicity measure
[26], and average magnitude difference function [27]. Another
important and widely used feature in this regard is signal
energy, which is presented in this work, and compared with the
dynamically calculated threshold.
IV. APNEA DETECTION USING VAD BASED - ENERGY
The general VAD block diagram used in our methodology
is shown in figure 1.
The assumptions on the VAD algorithm used here is that
the speech is quasi-stationary and its spectral changes quickly
over short periods like 20-30ms, but the background noise is
relatively stationary and changes very slowly with time. In
addition, the energy of the active speech level is usually higher
than background noise energy [28].
In the first step, the respiratory signal is filtered to remove
the undesired low frequency components. Then, the power with
different window sizes of the Fast Fourier Transform (FFT) is
calculated for the filtered signal [29].
Let x(t) be the input signal samples, and X(n) be Fast
Fourier Transform (FFT) samples. The VAD algorithm begins
with the energy computation within the smallest integer range
of frequency values n
1
and n
2
:
  


(1)
The energy of the signal is computed in two window
frames; short window and long window for every window
number i:
E
short
(i) = φ
short
E
nergy
+
(1 - φ
short
) E
short
(i) (2)
E
long
(i) = φ
long
E
nergy
+ (1 - φ
long
) E
long
(i) (3)
The number of frames used is N/L, where N represents the
number of samples in the signal and L represents the window
size in frequency domain. The coefficients φ
short
and φ
long
refer
to the window-length factors, where φ
short
= 1/16, and
φ
long
=
1/128, and L= 528 were used in this study.
At this point, since VAD aims to differentiate voice and
silence, where silence is mostly referring to background noise,
the noise level at every frame needs to be computed. For this
purpose, a threshold THR value needs to be determined for
comparing the signal value against noise:
THR =


+ M (4)
In the above formulation, K
f
is the K-th frame and M is a
margin value that can be considered to separate voice and
silence in the event that noise level is flat.
The VAD technique eventually makes a decision by
comparing every frame of signal energy against the THR value.
It is important to note that transitional periods from active
voice to silence may also affect the decision. Based on the
above steps and discussions, the decision on the VAD identifier
(ID) values is made as follows:
Figure 1. A block diagram of VAD design.
1, if Energy > THR
VAD- ID = 1, if Energy THR and in transitional period (5)
0, if Energy THR and not in transitional period
The outcome of the VAD technique is the separated speech
and silence phases which can be fine-tuned for identifying
breath versus breathing cessations for apnea detection.
In our work, there is a second threshold (Tr) that would be
used to decide whether a silence phase corresponds to apnea or
not. According to the sleep apnea literature, a breathing
cessation (silence) of 15 seconds or more would be classified as
apnea, as shown below:
If (VAD-ID
j
== 0) and (T
VAD-IDj
≥ 15Sec) =>
VAD-ID
j
is a SA event period
(6)
In the above relationship, T
VAD-IDj
corresponds to the
duration of silence phase j detected by the VAD technique.
The testing procedure and experimental results to evaluate
the described VAD algorithm will be discussed through the
next section.
V. EXPERIMENTAL RESULTS
MATLAB environment was used to perform our
methodology on various samples of breathing signals during
breathing and breath hold in 50 normal people. The volunteers
were asked to breath 20 cycles. They were asked to hold their
breath before, during and after the 20 cycles. The breath
recording was done using the SONY VAIO VPCEB42FM
microphone (Realtek High Definition Audio [30] professional
microphone with the accompanying Audacity software).
The human respiratory signal is given to the classification
system as the input, and the coding is developed in such way
that it calculates the fundamental feature of the respiratory
signal, which is the energy. The threshold is then applied to the
extracted energy feature and the binary decision is made.
VAD=1 is declared if the energy feature exceeds the threshold.
Otherwise, VAD=0 is for no breath or when silence (cessations
of breathing) is present.
Figure 2 shows the results obtained from the segmentation
technique of the input signal which splits the acoustic signal of
respiration into silence and voiced phases. The start point and
end point of a respiratory signal which contains breathing
Processing
Thresholds
Computation
VAD
decision
correction
VAD
decision
Features
Extraction
Input
Signal
VAD output
phases is determined in this work. Hence, the apnea events that
are silence phases lasting 15s or longer can be detected.
Figure 2. Segmentation acoustic signal of breath using VAD.
VI. CONCLUSIONS AND FUTURE WORK
This work sought to determine the effectiveness of VAD
based - energy in distinguishing the apnea in breathing signal.
The provided respiratory signal is classified successfully
with the help of the formulated algorithm with more than 97%
accuracy.
In order to detect sleep apnea in real time, the proposed
algorithm could be improved and adjusted by adding
calibration procedures to run on an FPGA [31]. After the
successful design and implementation of the OSA system, it is
planned to be experimentally tested in order to evaluate its
accuracy and practicality. The tests will take place in a local
hospital for a set of patients who have symptoms of OSA. In
addition, a different set of other subjects without SA symptoms
will test the system to verify its false positive accuracy.
REFERENCES
[1] Sleep Disorders Guide. www.sleepdisorderguide.com.
[2] N. Israel, A. Tarasiuk, and Y. Zigel, Nocturnal Sound Analysis for the
Diagnosis of Obstructive Sleep Apnea,” in Proceedings of the 32
nd
IEEE
International Conference on Engineering in Medicine and Biology
Society (EMBS 2010), pp. 6146-6149, Sep. 2010.
[3] A. Yilmaz, and T. Dundar, Home Recording for Pre-Phase Sleep Apnea
Diagnosis by Holter Recorder Using MMC Memory,” in Proceedings of
the 2010 IEEE International Conference on Virtual Environments
Human-Computer Interfaces and Measurement Systems (VECIMS), pp.
126-129, Sep. 2010.
[4] “Choosing a CPAP,” American Sleep Apnea Association,
www.sleepapnea.org/resources/pubs/cpap/htm.
[5] SleepMedInc. www.sleepmed.md.
[6] D. Baraglia et al., Automated Sleep Scoring and Sleep Apnea Detection
in Children,” in Proceedings of SPIE 6039, 2005.
[7] “Sleep Study Cost”, Shop & Compare Healthcare Facilities and Cost
with NewChoiceHealth.com. Web. 03 Dec. 2010.
www.newchoicehealth.com/SleepStudyCost.
[8] P. Chazal, C. Heneghan, and W. Mcnicholas, Multimodal Detection of
Sleep Apnoea Using Electrocardiogram and Oximetry Signals,”
Philosophical Transactions of The Royal Society A: Mathematical,
Physical and Engineering Sciences, vol. 367, no. 1887, pp. 369-389,
2009.
[9] L. Almazaydeh, M. Faezipour, and K. Elleithy, A Neural Network
System For Detection Of Obstructive Sleep Apnea Through SpO
2
Signal
Features,” (IJACSA) International Journal of Advanced Computer
Science and Applications, vol. 3, no. 5, pp. 7-11, Jun. 2012.
[10] L. Almazaydeh, K. Elleithy, and M. Faezipour, “Detection of obstructive
sleep apnea through ECG signal features,” in Proceedings of the IEEE
International Conference on Electro Information Technology (IEEE
eit2012), pp. 1-6, May. 2012.
[11] M. Canosa, E. Hernandez, and V. Moret, “Intelligent Diagnosis of Sleep
Apnea Syndrome,” In IEEE Engineering in Medicine and Biology
Magazine., vol. 23, no. 2, pp. 7281, 2004.
[12] P. Chazal, T. Penzel, and C. Heneghan, “Automated Detection of
Obstructive Sleep Apnoea at Different Time Scales Using the
Electrocardiogram,Institute of Physics Publishing., vol. 25, no. 4, pp.
967983, Aug. 2004.
[13] A. Yadollahi, and Z. Moussavi, “Acoustic Obstructive Sleep Apnea
Detection”, In Proceedings of IEEE Conference on Engineering in
Medicine and Biology Society (EMBC 2009), pp. 7110-7113, Sep. 2009.
[14] A. Ng, J. Chung, M. Gohel, et al., “Evaluation of the Performance of
Using Mean Absolute Amplitude Analysis of Thoracic and Abdominal
Signals for Immediate Indication of Sleep Apnoea Events,” Journal of
Clinical Nursing, vol. 17, no. 17, pp. 2360-2366, Sep 2008.
[15] Y. Zhao, H. Zhang, W. Liu, and S. Ding, “A Snoring Detector for
OSAHS Based on Patient’s Individual Personality,” In 3rd International
Conference in Awareness Science and Technology (iCAST)., pp. 24-27,
2011.
[16] K. Andrew, S. Tong, et al. Could Formant Frequencies of Snore Signals
Be an Alternative Means for the Diagnosis of Obstructive Sleep
Apnea?, Sleep Medicine, vol. 9, pp. 894-898, Dec. 2008.
[17] K. Andrew, T. Koh, E. Baey, et al., Speech-Like Analysis of Snore
Signals for the Detection of Obstructive Sleep Apnea, In International
Conference on Biomedical and Pharmaceutical Engineering 2006
(ICBPE 2006), pp. 99-103, Dec. 2006.
[18] T. Shochat, N. Hadas, M. Kerkhofs, et al., “The SleepStrip
TM
: An
Apnoea Screener for the Early Detection of Sleep Apnoea Syndrome,”
European Respiratory Journal, vol. 19, pp. 121-126, 2002.
[19] S. McClellan, J. Gibson,”Spectral entropy: An alternative indicator for
rate allocation,” In IEEE International Conference on Acoustics, Speech,
Signal Processing, pp. 201-204, Apr. 1994.
[20] B. Atal, L. Rabiner, “A Pattern Recognition Approach to Voiced-
Unvoiced-Silence Classification with Applications to Speech
Recognition,” IEEE Transactions on Acoustics, Speech, Signal
Processing, vol. 24, no. 3, pp. 201-212, June 1976.
[21] ETSI TS 126 094 V3.0.0 (2000-01), 3G TS 26.094 version 3.0.0 Release
1999, Universal Mobile Telecommunications System (UMTS);
Mandatory Speech Codec speech processing functions AMR speech
codec; Voice Activity Detector (VAD), 2000.
[22] A. Benyassine, E. Shlomot, and H. Su, “ITU-T recommendation G.729
annex B: A silence compression scheme for use with G.729 optimized
for V.70 digital simultaneous voice and data application,” IEEE
Communications Magazine, vol. 35, no. 9, 1997.
[23] E. Verteletskaya, K. Sakhnov, “Voice Activity Detection for Speech
Enhancement Applications,” ACTA Polytechnica, vol. 50, no. 4, pp. 100-
105, 2010.
[24] J. Haigh, and J. Mason, “Robust Voice Activity Detection Using
Cepstral Features,” In Proceedings of IEEE Region 10 Conference on
Computer, Communication, Control and Power Engineering
(TENCON’93), vol. 3, pp. 321-324, 1993, Beijing.
[25] A. Sangwan, C. MC, H. Jamadagni, et al., “VAD Techniques for Real-
Time Speech Transmission on the Internet,” In Proceedings of 5
th
IEEE
International Conference on High-Speed Networks and Multimedia
Communications, pp. 46-50, 2002.
[26] S. Tanyer, and H. Ozer, “Voice Activity Detection in Nonstationary
Gaussian Noise,” In Proceedings of International Conference on Signal
Processing ( ICSP’98), pp. 1620-1623, 1998.
[27] M. Orlandi, A. Santarelli, and D. Falavigna, “Maximum Likelihood
Endpoint Detection with Time Domain Features,” In Proceedings of 8
th
European Conference on Speech Communication and Technology
(eurospeech 2003), pp. 1757-1760, 2003, Geneva.
[28] K. Sakhnov, E. Verteletskay, and B. Simak, “Approach for Energy-
Based Voice Detector with Adaptive Scaling Factor,” IAENG
International Journal of Computer Science”, vol. 36, no. 4, pp. 394,
2009.
[29] A. Abushakra and M. Faezipour, “Acoustic Signal Classification of
Breathing Movements to Virtually Aid Breath Regulation,” IEEE
Journal of Biomedical and Health Informatics, vol. 17, no. 2, pp. 493-
500, March 2013.
[30] VAIO® Laptop ComputersVPCEB42FM/BJ: Available from:
http://www.docs.sony.com/release/VPCEB3_Series.pdf., 2011.
[31] B. Marinkovic, M. Gillette, and T. Ning, “FPGA Implementation of
Respiration Signal Classification Using a Soft-Core Processor,” In
Proceedings of the IEEE 31
st
Annual Northeast Bioengineering
Conference, pp. 54-55, April 2005.
... To automatically classify the respiratory signal to detect OSA, has elucidated on a self-developed algorithm. 4 Voice Activity Detection(VAD) has been used for classification of respiratory signals into positive or negative sleep apnea by checking for the apneic interval of fifteen seconds or more. Cepstral coefficient, spectral entropy, average magnitude difference function and signal energy are some of the important features of this algorithm. ...
... The distance to the human is first found out and then the breathing movements are tracked by performing in a shorter FFT and the reflected signals are monitored. Unlike papers 3,4 , this study also aims at detecting types of apneic intervals occurring in a subject. Threshold values are used in this study namely threshold on the minimum distance between two consecutive peaks and minimum amplitude at which a peak can be detected. ...
... Breathing pattern detection namely fine chest and abdomen movements due to breathing have given this study a better and stronger prediction of apnea event as compared to others. [3][4][5] Wearable chest bands having sensors fitted. These physiological signals can be monitored, thus being an important work in this particular domain. ...
... The percentage of correctly classified audio samples is referred to as accuracy [45]. It includes all four possibilities: True Positives (TP), True Negatives (TN), False Positives (FP), and False Negatives (FN). ...
Article
Full-text available
Genres are one of the key features that categorize music based on specific series of patterns. However, the Arabic music content on the web is poorly defined into its genres, making the automatic classification of Arabic audio genres challenging. For this reason, in this research, our objective is first to construct a well-annotated dataset of five of the most well-known Arabic music genres, which are: Eastern Takht, Rai, Muwashshah, the poem, and Mawwal, and finally present a comprehensive empirical comparison of deep Convolutional Neural Networks (CNNs) architectures on Arabic music genres classification. In this work, to utilize CNNs to develop a practical classification system, the audio data is transformed into a visual representation (spectrogram) using Short Time Fast Fourier Transformation (STFT), then several audio features are extracted using Mel Frequency Cepstral Coefficients (MFCC). Performance evaluation of classifiers is measured with the accuracy score, time to build, and Matthew's correlation coefficient (MCC). The concluded results demonstrated that AlexNet is considered among the top-performing five CNNs classifiers studied: LeNet5, AlexNet, VGG, ResNet-50, and LSTM-CNN, with an overall accuracy of 96%.
... Several studies in the topic of sleep research have been conducted, including detection of sleep apnea using, EEG signals [6], respiratory signals [7] and ECG signals [8]. Many of the existing works are computationally complex and require multiple physiological signals for initial screening of the disorder, which is not accessible in the rural health perspective. ...
... Bazı çalışmalarda ise horlama yerine solunum seslerine odaklanılmış, Almazaydeh ve ark 2013 solunum seslerinden apne tespiti yapmaya çalışmıştır. Epok tabanlı yaptıkları değerlendirmede apneli bölgeleri %97 oranında belirleyebildiklerini belirtmişlerdir [23]. ...
Article
Full-text available
Nowadays, increasing obesity and overweight due to eating and drinking habits cause many people to have sleep apnea syndrome, which manifests itself in the form of snoring, by obstructing the airways during sleep. Although there are diagnostic tools such as clinical and radiological examination, the gold standard measurement method for diagnosing this disease is PolySomnoGraphy (PSG). For the diagnosis of sleep apnea with polysomnography, while the patient is sleeping in the sleep laboratory in the hospital, many physiological parameters such as ECG, EEG, EMG, EOG, air flow in the mouth or nose, the chest and diaphragm circumference should be measured and evaluated by a specialist. Due to disadvantages such as patients who have difficulty sleeping in the hospital environment, small number of laboratories performing the test and the cost of the test, a new quest has begun. In this study, new methods developed as an alternative to polysomnography were reviewed and compared. As an alternative to polysomnography, Firstly, the methods based on measuring only a few physiological parameters in one's own home have been proposed. Many studies have been conducted to detect apnea by recording one or more of the respiratory signals, ECG, photoplethysmography, or pulse oximetry signals. These methods did not reach the expected clinical usage level due to the difficulties in the patient's self-enrollment. To overcome this problem, methods based on thermal camera or ultrasonic tracking, of the person that do not require electrical contact to the person have been proposed. As an alternative to these methods that require follow-up of the patient throughout the night, studies on apnea detection from speech voice recordings taken in the hospital environment for a few minutes while the person is awake have emerged. Similar study protocols were conducted in most of the articles reviewed. The methods generally include determination of features from the recorded signals, applying them to classifiers, deciding as to whether the person has apnea or the level of apnea. Among the alternative methods to polysomnography; it has been observed that apnea detection can be achieved with a success rate of over 90% with the methods that use physiological signal recordings during the person's sleep at home. On the other hand, a success of around 80% has been achieved in apnea detection studies from voice recordings made when the person was awake. As a result, it was evaluated that the detection of apnea from voice recordings of a few minutes while the person was awake would provide great convenience for the field, but its performance should be increased, and it should be confirmed by extensive clinical studies.
... The description of their denotation of calculation is tried on genuine respiratory symptom and result showing that the VAD algorithm is valuable as a prudent device for the division of breath into sound and quiet section. Apnea events may occur 5 to 30 times an hour and may occur up to four hundred times per night in those with severe SA [6]. E. Dafna et al. applied a novel system (method) in sleep characteristic analysis. ...
Conference Paper
This paper introduces an automated approach to identify the existence of Resistant Obstructive Sleep Apnea depending on breathing signal. The behavior of the breathing sounds is carried by decision making sound interval that detect respiratory signal when breathe and hold breathing. We have used the capability of recording breathing sound by a microphone and analyzed by sound frequency presence. Whether the recording data is abnormal has been analyzed with Audacity software. Respiratory signals are successfully analyzed by the decision making sound interval frequency made with 89.4% accuracy.
... Flexible temperature sensors are also used for monitoring breathing pattern 45 . Currently, the strain and 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 acoustics-based sensors often are used for monitoring breathing pattern 46,47 . However, the body movements during the measurements can limit the performances of such devices. ...
Article
Thin protein films of gelatin molecules grown on flexible substrates have been utilized to fabricate moisture-induced energy harvesting devices, which work as self-biased sensors. Adsorbed water molecules from the ambient moistures generate protons inside the film. A proton transfer path is formed through the hydrogen-bonded water molecules with protein around 55% relative humidity condition and the protons are transferred due to the gradient of absorbed water molecules within the protein films. The devices are capable to harvest electric power up to 5.5 μW/cm² with an induced voltage of 0.71V. Our findings not only provide a futuristic clean power generation concept from protein film as flexible power generator but also demonstrate the use of the energy harvesting devices as self-biased electronic sensors for various flexible and wearable applications. The devices showed exceptional performance as humidity sensors and have been used for flexible healthcare applications, such as, continuous monitoring of breathing pattern, lateral mapping of moisture levels at the finger tip for monitoring wound healing process. Nevertheless, the diode like response of the devices with humidity has been found to be suitable as self-biased humidity controlled electronic switch.
... Above all that studies [6], [7], [8] are based on signal combination: ECG and pulse oximetry for better results. Some studies do not use standard PSG signal at all, they are working with sound as the main source of information [9], [10]. The modern approach was selected by Ling Cen et al. [11], because they applied convolution neural network on three signals (saturation, airflow and abdominal movements). ...
... El ictus se caracteriza por un déficit neurológico a consecuencia de una lesión focal del sistema nervioso central de origen vascular. Existen diferentes factores de riesgo que predisponen a esta enfermedad [2], dentro de los cuales se encuentra el síndrome de apnea obstructiva del sueño (SAOS), el cual afecta aproximadamente al 2-4% de la población [3] y se caracteriza por el colapso total o parcial de la vía aérea superior durante el sueño, donde se presentan episodios de apnea e hipopnea que ocasionan una reducción en la saturación de oxígeno sanguíneo [3][4][5][6]. El SAOS toma relevancia en el ictus debido a que se relaciona con factores de riesgo cardiovasculares, como sobrepeso, edad avanzada, alcohol y tabaquismo [7]. ...
Article
Full-text available
Objetivo. Describir los principales mecanismos fisiopatológicos del síndrome de apnea obstructiva del sueño (SAOS) asociados con el desarrollo de ictus. Desarrollo. Los trastornos de la respiración del sueño tienen alta prevalencia en la población sana, entre los que el SAOS es el más reconocido. Este síndrome se ha asociado con enfermedades vasculares como el ictus, la hipertensión arterial y la fibrilación auricular, entre otras. El ictus posee una alta prevalencia mundial y se considera una enfermedad catastrófica. Los mecanismos fisiopatológicos son de gran importancia para comprender la relación que existe entre SAOS e ictus. Ambos trastornos se asocian con cambios moleculares, celulares y del sistema nervioso autónomo, con repercusiones cardiovasculares sistémicas y, en particular, con la salud vascular del cerebro. Conclusiones. La relación del SAOS y el ictus isquémico es evidente. El avance en la identificación de marcadores moleculares y en estudios de bajo coste para la identificación de SAOS permitirá implementar estrategias para reducir sus consecuencias en las enfermedades cardiovasculares, principalmente en el ictus isquémico. Palabras clave. Apnea del sueño. Fisiopatología. Ictus. Sueño. Trastornos de la respiración del sueño. Trastornos vasculares.
... El ictus se caracteriza por un déficit neurológico a consecuencia de una lesión focal del sistema nervioso central de origen vascular. Existen diferentes factores de riesgo que predisponen a esta enfermedad [2], dentro de los cuales se encuentra el síndrome de apnea obstructiva del sueño (SAOS), el cual afecta aproximadamente al 2-4% de la población [3] y se caracteriza por el colapso total o parcial de la vía aérea superior durante el sueño, donde se presentan episodios de apnea e hipopnea que ocasionan una reducción en la saturación de oxígeno sanguíneo [3][4][5][6]. El SAOS toma relevancia en el ictus debido a que se relaciona con factores de riesgo cardiovasculares, como sobrepeso, edad avanzada, alcohol y tabaquismo [7]. ...
Article
Aim: To describe the main physiopathological mechanisms of obstructive sleep apnea syndrome (OSAS) associated with the development of stroke. Development: Sleep breathing disorders have a high prevalence in the healthy population, among them, the OSAS is the most recognized. This syndrome has been associated with vascular diseases such as stroke, hypertension, atrial fibrillation, among others. Stroke has a high global prevalence and is considered a catastrophic disease. The physiopathological mechanisms are of great importance to understand the relationship that exists between OSAS and stroke. Both diseases are associated with molecular, cellular, and autonomic nervous system changes, with systemic cardiovascular repercussions and in particular with the vascular health of the brain. Conclusions: The relationship between OSAS and ischaemic stroke is evident. The advance in the identification of molecular markers and low-cost studies for the identification of OSAS will allow implementing strategies to reduce their consequences in cardiovascular diseases, mainly in ischaemic stroke.
... Another method developed in the literature for the separation of respiratory signals is the Voice Activity Detection (VAD) algorithm [20]. According to this method, breathholding and regular breath exchange are determined by the energy of the acoustic breathing signal. ...
Article
Obstructive sleep apnea is a disease that occurs in connection to pauses in respiration during sleep. Detection of the disease is achieved using a polysomnography device, which is the gold standard in diagnosis. Diagnosis is made by the steps of sleep staging and respiration scoring. Respiration scoring is performed with at least four signals. Technical knowledge is required for attaching the electrodes. Additionally, the electrodes are disturbing to an extent that will delay the patient’s sleep. It is needed to have systems as alternatives for polysomnography devices that will bring a solution to these issues. This study proposes a new approach for the process of respiration scoring which is one of the diagnostic steps for the disease. A machine-learning-based apnea detection algorithm was developed for the process of respiration scoring. The study used Photoplethysmography (PPG) signal and Heart Rate Variability (HRV) that is derived from this signal. The PPG records obtained from the patient and control groups were cleaned out using a digital filter. Then, the HRV parameter was derived from the PPG signal. Later, 46 features were derived from the PPG signal and 40 features were derived from the HRV. The derived features were classified with reduced machine-learning techniques using the F-score feature-selection algorithm. The evaluation was made in a multifaceted manner. Besides, Principal Component Analysis was performed to reduce system input (features). According to the results, if a real-time embedded system is designed, the system can operate with 16 PPG feature 95%, four PPG feature 93.81% accuracy rate. These success rates are highly sufficient for the system to work. Considering all these values, it is possible to realize a practical respiration scoring system. With this study, it was agreed upon that PPG signal may be used in the diagnosis of this disease by processing it with machine learning and signal processing techniques.
Article
Full-text available
Monitoring breath and identifying breathing movements have settled importance in many biomedical research areas, especially in the treatment of those with breathing disorders, e.g., lung cancer patients. Moreover, virtual reality (VR) revolution and their implementations on ubiquitous hand-held devices have a lot of implications, which could be used as a simulation technology for healing purposes. In this paper, a novel method is proposed to detect and classify breathing movements. The overall VR framework is intended to encourage the subjects regulate their breath by classifying the breathing movements in real time. This paper focuses on a portion of the overall VR framework that deals with classifying the acoustic signal of respiration movements. We employ Mel-frequency cepstral coefficients (MFCCs) along with speech segmentation techniques using voice activity detection and linear thresholding to the acoustic signal of breath captured using a microphone to depict the differences between inhale and exhale in frequency domain. For every subject, 13 MFCCs of all voiced segments are computed and plotted. The inhale and exhale phases are differentiated using the sixth MFCC order, which carries important classification information. Experimental results on a number of individuals verify our proposed classification methodology.
Conference Paper
Full-text available
Obstructive sleep apnea (OSA) is a common disorder in which individuals stop breathing during their sleep. Most of sleep apnea cases are currently undiagnosed because of expenses and practicality limitations of overnight polysomnography (PSG) at sleep labs, where an expert human observer is needed to work over night. New techniques for sleep apnea classification are being developed by bioengineers for most comfortable and timely detection. In this paper, an automated classification algorithm is presented which processes short duration epochs of the electrocardiogram (ECG) data. The automated classification algorithm is based on support vector machines (SVM) and has been trained and tested on sleep apnea recordings from subjects with and without OSA. The results show that our automated classification system can recognize epochs of sleep disorders with a high degree of accuracy, approximately 96.5%. Moreover, the system we developed can be used as a basis for future development of a tool for OSA screening.
Article
Full-text available
Obstructive sleep apnea (OSA) is a common disorder in which individuals stop breathing during their sleep. Most of sleep apnea cases are currently undiagnosed because of expenses and practicality limitations of overnight polysomnography (PSG) at sleep labs, where an expert human observer is needed to work over night. New techniques for sleep apnea classification are being developed by bioengineers for most comfortable and timely detection. In this paper, an automated classification algorithm is presented which processes short duration epochs of the electrocardiogram (ECG) data. The automated classification algorithm is based on support vector machines (SVM) and has been trained and tested on sleep apnea recordings from subjects with and without OSA. The results show that our automated classification system can recognize epochs of sleep disorders with a high degree of accuracy, approximately 96.5%. Moreover, the system we developed can be used as a basis for future development of a tool for OSA screening.
Article
Full-text available
Obstructive sleep apnea (OSA) is a common disorder in which individuals stop breathing during their sleep. These episodes last 10 seconds or more and cause oxygen levels in the blood to drop. Most of sleep apnea cases are currently undiagnosed because of expenses and practicality limitations of overnight polysomnography (PSG) at sleep labs, where an expert human observer is required. New techniques for sleep apnea classification are being developed by bioengineers for most comfortable and timely detection. In this study, we develop and validate a neural network (NN) using SpO2 measurements obtained from pulse oximetry to predict OSA. The results show that the NN is useful as a predictive tool for OSA with a high performance and improved accuracy, approximately 93.3%, which is better than reported techniques in the literature.
Conference Paper
Full-text available
The problem of voice activity detection in additive nonstationary Gaussian noise is considered. A new algorithm for the problem is presented. The algorithm utilizes the differences of the probability distribution properties of noise and speech signals. The magnitude density (mdf) and the magnitude distribution functions (MDF) are used to monitor the noise level for automatic threshold estimation. The estimate is shown to be accurate even when the analysis window does not fully contain non-speech signals and even in the presence of nonstationary noise. The voice activity detection algorithm is shown to operate reliably in SNR down to -5 dB. The method is compared with the periodicity measure method and zero-crossings method. Finally, a fusion algorithm utilizing those methods is suggested
Conference Paper
Full-text available
In this paper we propose an effective, robust and computationally low-cost HMM-based start-endpoint detector for speech recognisers . Our first attempts follow the classical scheme feature extractor-Viterbi classifier (used for voice activity detection) , followed by a post-processing stage, but the ultimate goal we pursue is a pure HMM-based architecture capable of performing the endpointing task. The features used for voice activity detection are energy and zero crossing rate, together with AMDF (Average Magnitude Difference Function), which proves to be a valid alternative to energy; further, we study the impact on performance of grammar structures and training conditions. In the end, we set the basis for the investigation of pure HMM-based architectures.
Article
This paper describes a study of noise-robust voice activity detection (VAD) utilizing the periodicity of the signal, full band signal energy and high band to low band signal energy ratio. Conventional VADs are sensitive to a variably noisy environment especially with low SNR, and also result in cutting off unvoiced regions of speech as well as random oscillating of output VAD decisions. To overcome these problems, the proposed algorithm first identifies voiced regions of speech and then differentiates unvoiced regions from silence or background noise using the energy ratio and total signal energy. The performance of the proposed VAD algorithm is tested on real speech signals. Comparisons confirm that the proposed VAD algorithm outperforms the conventional VAD algorithms, especially in the presence of background noise.
Conference Paper
A conventional diagnostic tool for assessing Obstructive Sleep Apnea Hypopnea Syndrome (OSAHS) is polysomnography (PSG), which is expensive and uncomfortable for patients. It is an important and urgent topic to find a non-invasive and low-cost diagnostic approach for OSAHS detection. Recently, the snore signal analysis receives much attention due to its potential capability for OSAHS detection. In this paper, we propose a novel method for diagnosing OSAHS based on patient's individual personality. First, the first formant frequencies of each snorer are classified into two clusters by K-means clustering. And then, using the first cluster center of each snorer, we set a personalized threshold to distinguish the hypopneic snores from the normal ones. Since the proposed threshold varies with each individual, the patient's individual personality can be overcome effectively. Experimental results show the validity of the proposed detector. In the experiments, the sensitivity of our method can achieve 90% and the specificity can achieve 91.67%.
Article
This paper investigates the automated detection of a patient's breathing rate and heart rate from their skin conductivity as well as sleep stage scoring and breathing event detection from their EEG. The software developed for these tasks is tested on data sets obtained from the sleep disorders unit at the Adelaide Women's and Children's Hospital. The sleep scoring and breathing event detection tasks used neural networks to achieve signal classification. The Fourier transform and the Higuchi fractal dimension were used to extract features for input to the neural network. The filtered skin conductivity appeared visually to bear a similarity to the breathing and heart rate signal, but a more detailed evaluation showed the relation was not consistent. Sleep stage classification was achieved with and accuracy of around 65% with some stages being accurately scored and others poorly scored. The two breathing events hypopnea and apnea were scored with varying degrees of accuracy with the highest scores being around 75% and 30%.
Conference Paper
Correlation between snoring and obstructive sleep apnea (OSA) has been studied over the past years using techniques such as sound intensity computation, power spectrum estimation, counting of snores, and more recently, pitch and jitter analysis. Although these studies have ascertained the relationship between snoring and OSA, they gave mixed outcomes in diagnosing OSA. In this paper, snore signals are analyzed using linear predictive coding (LPC), which is a powerful speech analytic tool. The noisy snore signals were first preprocessed using a Modified Normalized Least-Mean-Square (MNLMS) adaptive filter for noise cancellation, followed by the extraction of formant structures from the denoised snore signals to discriminate between benign and apneic snores. This initial study demonstrates that the formant structures of snore signals contain useful information for OSA detection, and suggests the feasibility of using snore signal analysis as a noninvasive and inexpensive diagnostic tool for mass screening of OSA.