Infant Cry Recognition Research Articles

Improving Infant Cry Recognition with CNNs and Imbalance Mitigation

Improving Infant Cry Recognition with CNNs and Imbalance Mitigation

Infant cry classification using CNN – RNN

The study of infant cry recognition aims to identify what an infant needs through her cry. Different crying sound can give a clue to caregivers about how to response to the infant’s needs. Appropriate responses on infant cry may influence emotional, behavioral, and relational development of infant while growing up. From a pattern recognition perspective, recognizing particular needs or emotions from an infant cry is much more difficult than recognizing emotions from an adult’s speech because infant cry usually does not contain verbal information. In this paper, we study the problem of classifying five different types emotion or needs expressed by infant cry, namely hunger, sleepiness, discomfort, stomachache, and indications that the infant wants to burp. We propose a novel approach using a combination of Convolutional Neural Networks (CNN) and Recurrent Neural Networks (RNN) that acts as feature extraction and classifier method at once. Particularly, CNN learns salient features from raw spectrogram information and RNN learns temporal information of CNN obtained features. We also apply 5-folds cross-validation on 200 training data set and 50 validation data set. The model with the best weight is tested on 65 test set. Evaluation in Dunstan Baby Language dataset shows that our CNN-RNN model outperforms the previous method by average classification accuracy up to 94.97%. The encouraging result demonstrates that the application of CNN-RNN and 5-folds cross-validation offers accurate and robust result.

A novel infant cry recognition system using auditory model‐based robust feature and GMM‐UBM

SummaryRecognizing infant cry is a meaning work, which can help new parent to understand infant's needs. Mostly, the motivation of existed recognition features is based on the psychoacoustic model. However, weak features are not enough to represent the details of infant cry. To address this issue, we propose a novel infant cry recognition system. In our system, the feature extraction method derived from auditory model, this model can address the auditory neural active representation. Additionally, we designed the intelligent system by a classifier named Gaussian Mixture Model‐Universal Background Model, which has a robust recognition performance to channel imbalance and corroded signal. The experiment results have shown a superior performance. Compared with the High‐order Spectral feature, a state‐of‐the‐art feature, when using Gaussian Mixture Model and General Background Model as classifier, the recognition accuracy improved by 14.6% and 13.7% under clear and corroded signal, respectively.

Infant cry language analysis and recognition: an experimental approach

Recently, lots of research has been directed towards natural language processing. However, the baby &#x02BC s cry, which serves as the primary means of communication for infants, has not yet been extensively explored, because it is not a language that can be easily understood. Since cry signals carry information about a babies &#x02BC wellbeing and can be understood by experienced parents and experts to an extent, recognition and analysis of an infant &#x02BC s cry is not only possible, but also has profound medical and societal applications. In this paper, we obtain and analyze audio features of infant cry signals in time and frequency domains. Based on the related features, we can classify given cry signals to specific cry meanings for cry language recognition. Features extracted from audio feature space include linear predictive coding &#x0028 LPC &#x0029, linear predictive cepstral coefficients &#x0028 LPCC &#x0029, Bark frequency cepstral coefficients &#x0028 BFCC &#x0029, and Mel frequency cepstral coefficients &#x0028 MFCC &#x0029. Compressed sensing technique was used for classification and practical data were used to design and verify the proposed approaches. Experiments show that the proposed infant cry recognition approaches offer accurate and promising results.

Development of Portable Medical Electronic Device for Infant Cry Recognition

Infant cry is considered as the first biological signal communicated by the newborns to the mother and the outside world. In NICU, there is a huge need to monitor the physiological conditions of the premature and full term infants. In such environment, the cry signal conveys significant information in terms of the distress of the infant and thereby needs of the infants can be well attended. Establishment of the portable device is thus essential for such scenario. This work enumerates a primitive experimental study on developing portable electronic medical device for infant cry recognition. The setup comprises of a voice recorder, NI My RIO processor with a Lab view followed by the LCD Tablet for display. The cry signals were recorded by placing the voice recorder in the incubator and were qualitatively assessed by the clinician and confirm its suitability for the experimental study. A threshold based classification was employed which sends the voice based alarm to the tablet. The continuous monitoring of such facility in NICU provides the behavioral status of the infant

Application of Pattern Recognition Techniques to the Classification of Full-Term and Preterm Infant Cry

Scientific and clinical advances in perinatology and neonatology have enhanced the chances of survival of preterm and very low weight neonates. Infant cry analysis is a suitable noninvasive complementary tool to assess the neurologic state of infants particularly important in the case of preterm neonates. This article aims at exploiting differences between full-term and preterm infant cry with robust automatic acoustical analysis and data mining techniques. Twenty-two acoustical parameters are estimated in more than 3000 cry units from cry recordings of 28 full-term and 10 preterm newborns. Feature extraction is performed through the BioVoice dedicated software tool, developed at the Biomedical Engineering Lab, University of Firenze, Italy. Classification and pattern recognition is based on genetic algorithms for the selection of the best attributes. Training is performed comparing four classifiers: Logistic Curve, Multilayer Perceptron, Support Vector Machine, and Random Forest and three different testing options: full training set, 10-fold cross-validation, and 66% split. Results show that the best feature set is made up by 10 parameters capable to assess differences between preterm and full-term newborns with about 87% of accuracy. Best results are obtained with the Random Forest method (receiver operating characteristic area, 0.94). These 10 cry features might convey important additional information to assist the clinical specialist in the diagnosis and follow-up of possible delays or disorders in the neurologic development due to premature birth in this extremely vulnerable population of patients. The proposed approach is a first step toward an automatic infant cry recognition system for fast and proper identification of risk in preterm babies.

Right Frontoinsular Cortex and Subcortical Activity to Infant Cry Is Associated with Maternal Mental State Talk.

The current study advances extant literature examining the neural underpinning of early parenting behavior. The findings highlight the special functional importance of the right frontoinsular cortex-thalamic-limbic network in a mother's proclivity to engage in mental state talk with her preverbal infant, a circumscribed aspect of maternal caregiving purported to be a prerequisite of sensitive and responsive caregiving. These associations existed specifically for maternal mentalizing behavior and were not evident for more generic aspects of caregiving in this urban sample of 76 postpartum mothers. Finally, the findings were robust even when controlling for potential demographic, perceptual, and contextual confounds, supporting the notion that these regions constitute an innate, specialized maternal mentalizing network.

An Arduino-Based Resonant Cradle Design with Infant Cries Recognition.

This paper proposes a resonant electric cradle design with infant cries recognition, employing an Arduino UNO as the core processor. For most commercially available electric cradles, the drive motor is closely combined with the bearing on the top, resulting in a lot of energy consumption. In this proposal, a ball bearing design was adopted and the driving force is under the cradle to increase the distance from the object to fulcrum and torque. The sensors are designed to detect the oscillation state, and then the force is driven at the critical time to achieve the maximum output response while saving energy according to the principle of resonance. As for the driving forces, the winding power and motors are carefully placed under the cradle. The sensors, including the three-axis accelerometer and infrared sensor, are tested and applied under swinging amplitude control. In addition, infant cry recognition technology was incorporated in the design to further develop its functionality, which is a rare feature in this kind of hardware. The proposed nonlinear operator of fundamental frequency () analysis is able to identify different types of infant cries. In conclusion, this paper proposes an energy-saving electric cradle with infant cries recognition and the experimental results demonstrate the effectiveness of this approach.

Automatic Recognition System of Infant Cry based on F-Transform

A new approach has been proposed to design of an automatic infant cry recognition system based on the fuzzy transform (F-transform) that classifies two different kinds of cries, which come from physiological status and medical disease. Feature extraction and development of classification algorithms play an important role in the area of automatic analysis of infant cry signals. F-transform is the powerful tool for the approximation of continuous functions on a finite domain and has been used as feature extraction of infant cry signals. Neural networks is developed to classify the infant cry signals into physiological status and medical disease and trained with smoothing parameter to obtain better classification accuracy. Neural networks (recognition system) are trained from the random selection of crying babies. The experimental results show that the proposed give very promising classification accuracy of 96%. The proposed method can be used to assist medical investigation of the infant from the cry signals.

Classification of Infant Cries Using Dynamics of Epoch Features

AbstractIn this article, epoch-based dynamic features such as sequence of epoch interval values and epoch strength values are explored to classify infant cries. Epoch is the instant of significant excitation of the vocal tract system during the production of speech. For voiced speech, the most significant excitation takes place around the instant of glottal closure. The different types of infant cries considered in this work are hunger, pain, and wet diaper. In this work, epoch strength and epoch interval features are used to represent infant cry-specific information from the acoustic signal. In this study, the proposed features such as epoch interval and epoch strength values are determined using zero-frequency filter-based method. Gaussian mixture models (GMMs) are used to classify the above-mentioned cries from the features proposed in this work. GMMs are developed separately for each of the cries using the proposed features. The infant cry database collected under a telemedicine project at the Indian Institute of Technology Kharagpur has been used for this study. In the first step, infant cry recognition accuracy is investigated separately using epoch interval and epoch strength features. To enhance recognition performance, GMMs developed using various features are combined through score level fusion techniques. The recognition performance using a combination of evidence is found to be superior over individual systems.

Optimization of Principal Component Analysis and Support Vector Machine for the Recognition of Infant Cry with Asphyxia

The optimization of principal component analysis–support vector machine (PCA–SVM) for recognizing infant cry with asphyxia is presented in this paper. Three types of PC A selection techniques such as cumulative percent of variance, eigenvalue-one-criterion and scree test were employed to select significant features of Mel-frequency cepstrum coefficient that is extracted from normal and asphyxia cry. The asphyxiated infant cries were differentiated from normal cries using SVM with linear and radial basis function (RBF) kernels. The performance of the PCA–SVM in recognizing asphyxiated infant cries was compared with the SVM (without PCA) to prove its efficiency. Classification accuracy and support vector number were computed to examine the performance of both techniques. The results show that PCA–SVM is the best technique for recognizing asphyxiated infant cries since it produces the highest classification accuracy (95.323%). The RBF kernel with optimal regularization of 100 and γ=0.025 should be used in the PCA–SVM technique.