|Year : 2022 | Volume
| Issue : 3 | Page : 410-415
Baby crying analyzing and solution using matlab graphical user interface; interdisciplinary collaboration between engineering and nursing
Efe Çetin Yilmaz1, Serap Ozdemir2
1 Department of Control Systems Electrical and Electronic Engineering, Faculty of Engineering and Architecture, Kilis 7 Aralik University, Kilis, Turkey
2 Department of Nursing, Faculty of Health Sciences, Kilis 7 Aralik University Yusuf Serefoglu, Kilis, Turkey
|Date of Submission||03-Jun-2022|
|Date of Decision||15-Jul-2022|
|Date of Acceptance||15-Aug-2022|
|Date of Web Publication||17-Sep-2022|
Department of Nursing, Faculty of Health Sciences, Kilis 7 Aralik University Yusuf Serefoglu, Kilis
Source of Support: None, Conflict of Interest: None
Background: Babies can express all their needs (such as hunger, pain, tiredness, discomfort, and so on) to their parents with crying behavior that being able to predict these behaviors of babies correctly parents is extremely important for the comfort of babies. In recent years, analyzing the baby crying sound and interpreting it in line with the needs has been developing as an important process in the estimation of baby needs. Methods: Analyzing the spectra of the baby crying sound over time and amplitude period creates a significant knowledge base on the prediction of baby needs. Within the scope of this study, a new method has been developed for the development of various technical analyzes of a sample baby crying sound using the MATLAB program. Results: With this method, the energy fluctuations in the sample baby crying sound were analyzed, and the changes in the crying process were examined through the baby crying process. Conclusions: As a result, thanks to the analysis data obtained within the scope of this study, it is aimed to provide data to autonomous controlled baby care units that can be manufactured in future studies.
Keywords: Artificial intelligent, baby crying analyses, MATLAB modeling, Simulation
|How to cite this article:|
Yilmaz EÇ, Ozdemir S. Baby crying analyzing and solution using matlab graphical user interface; interdisciplinary collaboration between engineering and nursing. Biomed Biotechnol Res J 2022;6:410-5
|How to cite this URL:|
Yilmaz EÇ, Ozdemir S. Baby crying analyzing and solution using matlab graphical user interface; interdisciplinary collaboration between engineering and nursing. Biomed Biotechnol Res J [serial online] 2022 [cited 2022 Dec 8];6:410-5. Available from: https://www.bmbtrj.org/text.asp?2022/6/3/410/356150
| Introduction|| |
Various artificial intelligence maps are created in the analysis of living tissue behavior in computer-aided programs, and methods are developed for the analysis of behavior., The act of crying in newborn babies is a complex neurophysiological condition. Crying occurs in response to a particular stimulus and causes perinatal stress, hunger, pain, anger, sleepiness, etc., in the baby. Cry response is mainly regulated by structural or functional changes in neurological areas such as the hypothalamus, amygdala, caudal periaqueductal region, and cranial nerves. The act of crying appears to involve energy-producing metabolism, including psychoneurological, respiratory, musculoskeletal, cardiovascular, and genetic involvements. Disturbances in one or more of these markers may be responsible for an imperfect crying process. Crying peaks at around 6 weeks and decreases by 4 months and then remain stable. The type, frequency, and duration of crying are highly variable, especially after early infancy. It is a known fact that babies report all their needs with crying behavior. It is often very difficult for those responsible for the baby to determine the needs of the baby with this crying behavior. It is known that experienced people can understand the needs of babies from the way they cry and act accordingly, but an individual's differences and some complex situations can make this difficult. It may be easier for parents to recognize the needs and problems of the baby in the 1st days of life., However, it is not always easy to make these analyzes by a machine. The recent massive development of measuring instruments and analysis software and larger sample sets has begun to contradict the belief that the infant's needs cannot be determined. Studies suggest that facial expressions and crying features can reliably distinguish between cries of pain, hunger, sadness, fear, and anger.
In recent years, it has been possible to detect a baby's crying sound and to recognize the baby's needs thanks to the latest technologies developed (baby cry transfer) that classified baby crying sounds with the help of an embedded device. After this device recorded the sounds through an external microphone, it divided the reason for crying into pain, bedwetting, open and other types. In the study of the microphone-connected device performed both recording and processing of crying signals (signal detection sound activity detection voice activity detection algorithm was used) through Raspberry Pi, and it was determined that the reasons for crying were hunger, pain, or fever. In literature, Priscilla Dunstan found that first trimester babies use proto-language to communicate, that is, five words to express their needs. These five words are “Neh” (hungry), “Eh” (need to pass gas), “Owh/Oah” (fatigue), “Eair/Eargghh” (cramps), and “Heh” (physical discomfort and feeling hot or wet)., In the study of the relationship between the sound recordings taken from babies and the disease was examined, and when 0–7-month-old babies were compared, normal and asphyxia (deprived of oxygen) babies were classified. In the analysis of the study, the effect of the least squares method on the classification performance of multilayer perceptrons (MLPs) is discussed. As a result, 94% accuracy performance has been achieved by reducing the computational load of this method and increasing the accuracy of multilayer sensors. In the study, the voices of 0–6 wet babies were recorded 24/7 in the home environment. Two different machine learning algorithms as logistic regression and convolutional neural networks (CNNs) have been compared, and it has been reported that CNNs are more advantageous. It is of great importance for the comfort of the baby that the parents can accurately predict the crying behavior and needs of the baby. As a result of the studies, it has been reported that the crying analyzes of healthy and sick babies of different models are effective in determining their needs and providing ease of care for the family. It is seen that there is a need to create new models to expand these models and obtain the most accurate results. For this reason, in this study, the signals occurring in a sample baby crying process were analyzed in frequency and amplitude modulations using the MATLAB program.
| Methods|| |
In this study, the analysis of a sample produced on the program vector-based newborn baby crying sound was carried out under the conditions determined on the MATLAB R2019 A version base as computer analyses method program. In the analysis process, the baby crying sound was first converted to i-vector graphics. [Figure 1] shows the MATLAB program worksheet used in this study and the vector analysis of the sample baby crying sound. Converting the baby crying sound to vector graphics enabled the sound formed during the analysis process to be evaluated in terms of both amplitude and intensity. Magnitude sum frequency (MSF) of the signal analyzed in this study can be taken as the average absolute value of the basic input signal and expressed by the following formula.
|Figure 1: MATLAB program worksheet used in this study and the vector analysis of the sample baby crying sound|
Click here to view
In this process, using individual signals while analyzing this feature of the signal will result in a noisy MSF signal. This will cause many problems in the analysis of the signal. Therefore, in this study, we use MSF as a combination of pitch and ZC in fundamental signal analysis. In real-time sound analysis, the pitch can be defined as a subjective psychoacoustic feature. This analysis contributes to the analysis of the basic frequency of the signal that occurs during the baby's crying. In a mature human voice analysis, the pitch frequency can be obtained in the range of 100–350 Hz, while in a real-time baby crying behavior, the pitch frequency can be obtained in the range of 350–500 Hz. In this study, it was used to differentiate the baby's crying behavior from talking and noise. In this way, real-time baby crying sound was filtered from noise and speech frequencies. [Figure 2] shows the time axis behavior and amplitude spectrum of the real-time baby crying sound analyzed in this study, respectively.
|Figure 2: The time axis behavior and amplitude spectrum of the real-time baby crying sound analyzed in this study respectively. (a) frequency distribution in the time band, (b) amplitude distribution in time band|
Click here to view
It is possible to define the autocorrelation function as follows,
In this function, considering the input as a periodic signal, the total range ranges from summation (–) inf to (+) inf,
Here, AC can be defined as function equality in general terms.
We can define the STAC function for this process as follows
The parameter n that composes this equation is the length of the window, w represents the defined function, and m represents the current sampling signal. Considering eqn.(4)
If we define Then,
Kth AC at time n is obtained by filtering × (n) × (n-k). A filter with an impulse response will correspond to the spectrum of the sampled signal. Thus, the real-time baby crying sound will be analyzed in the time axis with the pitch method. According to the results of the analysis, baby behaviors (for example, hungry, pain, sleep, etc.) can be predicted according to the pitch changes in the baby crying sound.
| Results|| |
In this study, frequency base and spectrum analyzes of the real-time baby crying sound using a computer analyses program (MATLAB) are shown in [Figure 3]. In this analysis, it is seen that the baby's crying sound is generally clustered around a circle. This shows that in the spectrum analysis of baby crying, it is necessary to focus on the elevation values in peak signal regions. It is of great importance to catch the key sounds that babies need in the propagation of rising signals. With the analysis model performed in this study, the peak values in the baby crying sound will be analyzed by different methods and verified. In this way, artificial intelligence maps were created to estimate the baby's needs more accurately. In addition, the age and sex changes of babies will be analyzed with the changes to be made on the program in living tissue studies.
|Figure 3: Frequency base and spectrum analyzes of real-time baby crying sound using computer analyses program (MATLAB). (a) sample frequency signal, (b) Distribution of the signal in the solution set, (c) amplitude distribution on peak in time band|
Click here to view
[Figure 4] shows the spectrum analysis of the sample baby crying sound signal in the frequency band in this study. According to this analysis, the distribution of peak values occurring in the real-time baby crying sound will be analyzed based on MATLAB, so that its overlap with similar sounds in the live environment can be analyzed throughout the baby crying process. In literature, it is when these spectrums are examined that the differences between the baby crying spectrum and the human speech spectrum will be seen in the frequency domain. There are similarities and differences between the adult voice and the voice of crying babies. In previous literature studies, differences in the character of the voice were found in terms of the basic frequency (pitch) where the voice of crying babies is higher.
|Figure 4: Spectrum analysis of the sample baby crying sound signal in the frequency band|
Click here to view
| Discussion|| |
In recent years, the use of various analyzes in the field of biotechnology using computer-aided programs is frequently preferred method. With these analyzes, samples taken from living tissue are processed on the artificial intelligence map and some prediction mechanisms are tried to be created. This prediction mechanism can be used for the needs of the living thing or for a treatment method. Over the past years, numerous and different efforts and studies have been conducted on infants over the generations seeking to establish general laws that describe the process of formation and perception of baby crying not only as an acoustic linguistic but also as an indicator of neurophysiological state. Increasing the number of samples in experimental sets will greatly contribute to the interpretation of the signal. In a study in the literature, it was reported that 182 crying units from the Mexican baby set and 65 units from the Cuban baby set were detected with the determining method. The physiological and physical similarities of the babies included in the experimental cluster will contribute to the formation of signals in more similar clusters. Therefore, performing a standard clustering of infants in real-time infant cry analysis studies will increase the given accuracy.
Newborn baby crying is generally caused by the rhythmic transitions between inhalation and exhalation, where the vibration of the vocal cords produces periodic air pulses. It is assumed that the average value of these vibration pulses in infants is between 250 and 600 Hz. As a result, the resulting cry signal is shaped by the vocal tract and is released, causing resonant frequencies. The first two formants typically occur around 1100 Hz and 3300 Hz, respectively. Some studies indicate that early in life people exhibit a variety of crying patterns: protest crying (like being left in a cradle, where the baby is faced with loss and wants to take back the loss) and the sad cry of despair (a low cry indicating acceptance of the loss)., Sometimes, additional types such as hunger and painful crying are also considered. In recent years, the analysis of babies' crying and what these crying findings express have gained importance. It is stated that deep learning methods can be applied for successful automatic classification. Automatic methods are used with classical approaches such as a newborn baby crying Fourier transform and autocorrelation analysis and parametric techniques. These methods allow the estimation of the main acoustic properties such as vocal fold vibration frequency, vocal tract resonance frequencies, crying duration, and the like. Crying analyzes are examined on healthy and sick babies in line with certain models. In literature, collected and processed real-time crying signals of babies using the Intelligent Cry Detection System and Raspberry Pi 3. The notification, in which the system classified the reason for crying in babies, sent an SMS to the families through the mobile phone through Wi-Fi. In this context, it offers approaches to emergency or nonemergency situations through notifications. It has been reported to families that these notifications provide a great convenience in meeting their baby care needs. In the study of an application was developed to meet the needs of the babies of hearing-impaired and healthy families. Using machine learning methods, both voice and speech systems were used. As a result of the application, the baby's needs are reported on the screen in writing and audibly, and it is stated that it provides a lot of convenience for families with disabilities.
[Figure 4] shows the spectrum analysis of the sample baby crying sound signal in the frequency band in this study. According to this analysis, the distribution of peak values occurring in the real-time baby crying sound will be analyzed based on MATLAB, so that its overlap with similar sounds in the live environment can be analyzed throughout the baby crying process. In literature, it is when these spectrums are examined that the differences between the baby crying spectrum and the human speech spectrum will be seen in the frequency domain. There are similarities and differences between the adult voice and the voice of crying babies. In previous literature studies, differences in the character of the voice were found in terms of the basic frequency (pitch) where the voice of crying babies is higher. However, the analysis and integration of this spectrum into the system have not yet been clarified by the researchers. The voice of crying babies has short vocal cords and has regular characters as can be seen from the spectrogram because it is thin. Priscilla Dunstan proposed the idea of determining the meaning of a crying baby named Dunstan Baby Language Crying babies have five types of universal sounds and their meanings are as follows:
- ”Neh”: The “new” sound comes from sucking and pushing the tongue into the mouth, which means the baby is hungry
- ”Owh/Oah”: His voice “Owh” sounds like a man yawning, which means the baby is sleeping
- ”Heh”: The “Heh” sound is derived from the baby's response to burning or itching, which means that the baby is not comfortable
- ”Earth/Eargghh”: An “Eairh” sound occurs when the baby does not have burping which causes air bubbles to enter his stomach and cannot be released, which means the baby has stomach problems
- ”Eh”: The “eh” sound occurs when the wind is trapped and not out of the chest, causing air to come out of the mouth, meaning the baby wants to burp.
The target cluster determined by the increase in the baby cry sample sound in the time band will be more sensitive. Thus, the deviation between different sounds can be analyzed with computer-aided programs, and more confirmatory equations can be created on the estimation of newborn needs. In future studies, it is seen that the predictive ability of computer-aided programs will be greater as the sound samples taken in the needs clustering on real-time newborn cry are increased in test procedures. In the signal processing model obtained within the scope of this study, it is estimated that the accuracy rate can be over 80% with a sufficient number of samples. In literature, present results on the classification of neonatal crying as a normal and hypoxia-related disorder using radial basic function neural networks with 85% overall classification performance. Sahak et al. combined support-vector machine and principal component analysis were applied to recognize asphyxial baby cries with 95.86% classification accuracy. In literature, another study applied a new algorithm to optimize Mel-frequency cepstrum coefficients to extract an optimal feature set for diagnosing hypothyroidism in infants using an MLP neural network. The algorithm provides classification accuracy of up to 95% (average 86%) to identify expiratory and inspiratory phases from baby cries. In another study in literature by Hariharan et al., a general regression neural network was used as a classifier to distinguish between normal crying signals and pathological crying signals from deaf infants and infants with asphyxia.
| Conclusions|| |
It is not always an easy process to determine the crying behavior of newborn babies and their needs. It is thought that the ability to accurately predict the crying behavior of babies and their needs will contribute significantly to families and health professionals. Therefore, the use of signal processing technology in baby crying analyzes will contribute significantly to the estimation of needs. The engineering solutions realized in this area have a great impact on the comfort of the baby and the family. However, the stability and accuracy of the designed systems will be possible with interdisciplinary studies. Therefore, the application of the designed engineering solutions in the in vivo environment and the control of the results will be of great importance in terms of system development. As a result, it is predicted that determining the needs of newborn babies, whose only communication tool is crying, with these analyzes will contribute significantly to the quality of treatment care services. In future studies, it is important to examine analyzing the system outputs of the designed engineering solutions for babies crying of different months and genders.
Limitation of the study
The limitation of the current study is to analyze the sample baby crying sound but not to increase the sample variables.
Financial support and sponsorship
Conflicts of interest
The authors declare that none of the authors have any competing interests.
| References|| |
Sabeenian R, Vijitha V. Identification and categorization of brain tumors using ensemble classifiers with hybrid features. Biomed Biotechnol Res J (BBRJ) 2021;5:357-65.
Ozdemir S, Yilmaz E. Interdisciplinary collaboration between engineering and nursing on baby crying analyzing and classification: A biotechnology study. J Prev Diagn Treat Strategies Med 2022;1:82-7. [Full text]
Bashiri A, Hosseinkhani R. Infant crying classification by using genetic algorithm and artificial neural network. Acta Med Iran 2020;58:531-9.
Owings D, Zeifman D. Human infant crying as an animal communication system: Insights from an assessment/management approach. In: Oller D, Griebel U, editors. Evolution of Communication Systems: A Comparative Approach. Cambridge, MA: MIT Press; 2004. p. 151-70.
Senese VP, Azhari A, Shinohara K, Doi H, Venuti P, Bornstein MH, et al.
Implicit associations to infant cry: Genetics and early care experiences influence caregiving propensities. Horm Behav 2019;108:1-9.
Richey L, Li T, Rilling JK. Perception of male and female infant cry aversiveness by adult men. J Reprod Infant Psychol 2022;40:76-90.
Pal A, Iyer N, Yantorno RE. Emotion detection from infant facial expressions and cries. IEEE Int Conf Acoust Speech Signal Process Proc 2006;2:2-3.
Saha B, Purkait PK, Mukherjee J, Majumdar AK, Majumdar B, Singh AK. An Embedded System for Automatic Classification of Neonatal Cry. 2013 IEEE Point-of-Care Healthcare Technologies (PHT); 2013. p. 248-51.
Patil MS, Kamerikar UA. Cause estimation of neonatal cry using raspberry pi. Int J Appl Eng Res 2018;13:10722-6.
Franti E, Ispas I, Dascalu M. Testing the Universal Baby Language Hypothesis-Automatic İnfant Speech Recognition with CNNs. 41st
International Conference on Telecommunications and Signal Processing (TSP) IEEE; 2018. p. 1-4.
Zabidi A, Khuan LY, Mansor W, Yassin IM, Sahak R. Classification of İnfant Cries with Asphyxia Using Multilayer Perceptron Neural Network. 2010 Second International Conference on Computer Engineering and Applications; 2010. p. 204-8.
Cohen R, Ruinskiy D, Zickfeld J, IJzerman H, Lavner Y. Baby cry detection: Deep learning and classical approaches. In: Development and Analysis of Deep Learning Architectures. Cham: Springer; 2020. p. 171-96.
Rúız MA, Reyes CA, Altamirano LC. On the implementation of a method for automatic detection of infant cry units. Procedia Eng 2012;35:217-22.
Jagtap SS, Kadbe PK, Arotale PN. System Propose for be Acquainted with Newborn Cry Emotion Using Linear Frequency Cepstral Coefficient. In: 2016 International Conference on Electrical, Electronics, and Optimization Techniques (ICEEOT). Chennai, India: IEEE; 2016. p. 238-42.
LaGasse LL, Neal AR, Lester BM. Assessment of infant cry: Acoustic cry analysis and parental perception. Ment Retard Dev Disabil Res Rev 2005;11:83-93.
Hendriks MC, Nelson JK, Cornelius RR, Vingerhoets AJ. Why crying improves our well-being: An attachment-theory perspective on the functions of adult crying. In: Emotion Regulation. Boston, MA: Springer; 2008. p. 87-96.
Viragova M, O'Curry S. Understanding persistent crying in infancy. Paediatr Child Health 2021;31:116-21.
Kheddache Y, Tadj C. Acoustic measures of the cry characteristics of healthy newborns and newborns with pathologies. J Biomed Sci Eng 2013;6:796-803.
Myakala PR, Nalumachu R, Sharma S, Mittal VK. An İntelligent System for İnfant Cry Detection and İnformation in Real Time. 2017 Seventh International Conference on Affective Computing and Intelligent Interaction Workshops and Demos (ACIIW). San Antonio; 2017. p. 141-6.
Bütüner R. Estimation of Baby Sounds with Artificial Intelligence Methods. J Inf Syst Manage Res 2020;2:32-41.
Dewi SP, Prasasti AL, Irawan B. The Study of Baby Crying Analysis Using MFCC and LFCC in Different Classification Methods. In: 2019 IEEE International Conference on Signals and Systems (ICSigSys). Bandung, Indonesia: IEEE; 2019. p. 18-23.
Franti E, Ispas I, Dascalu M. Testing the Universal Baby Language Hypothesis-Automatic İnfant Speech Recognition with CNNs. In: 2018 41st
International Conference on Telecommunications and Signal Processing (TSP) Athens, Greece. IEEE; 2018. p. 1-4.
Poel M, Ekkel T. Analyzing infant cries using a committee of neural networks in order to detect hypoxia related disorder. Int J Artif İntell Tools 2006;15:397-410.
Sahak R, Mansor W, Lee YK, Yassin AM, Zabidi A. Performance of combined support vector machine and principal component analysis in recognizing infant cry with asphyxia. Annu Int Conf IEEE Eng Med Biol Soc 2010;2010:6292-5.
Hariharan M, Sindhu R, Yaacob S. Normal and hypoacoustic infant cry signal classification using time-frequency analysis and general regression neural network. Comput Methods Programs Biomed 2012;108:559-69.
[Figure 1], [Figure 2], [Figure 3], [Figure 4]