Abstract
Crying in babies is a primary communication function, governed directly by the brain; any alteration on the normal functioning of the babies’ body is reflected in the cry (Wasz-Höckert, et al, 1968). Based on the information contained in the cry’s wave, the infant’s physical state can be determined; and even pathologies in very early stages of life detected (Wasz-Höckert, et al, 1970). To perform this detection, a Fuzzy Relational Neural Network (FRNN) is applied. The input features are represented by fuzzy membership functions and the links between nodes, instead of weights, are represented by fuzzy relations (Reyes, 1994). This paper, as the first of a two parts document, describes the Infant Cry Recognition System´s architecture as well as the FRNN model. Implementation and testing are reported in the complementary paper.
TopBackground
The pioneer works on infant cry were initiated by Wasz-Hockert since the beginnings of the 60’s. In one of those works his research group showed that the four basic types of cry can be identified by listening: pain, hunger, pleasure and birth. Further studies led to the development of conceptual models that describe the anatomical and physiologic basis of the production and neurological control of crying (Bosma, Truby & Antolop, 1965). Later on, Wasz-Hockert (1970) applied spectral analysis to identify several types of crying. Other works showed that there exist significant differences among the several types of crying, like healthy infant’s cry, pain cry and pathological infant’s cry. In one study, Petroni used Neural Networks (Petroni, Malowany, Johnston, and Stevens, 1995) to differentiate between pain and no-pain crying. Cano directed several works devoted to the extraction and automatic classification of acoustic characteristics of infant cry. In one of those studies, in 1999 Cano presented a work where he demonstrates the utility of the Kohonen’s Self-Organizing Maps in the classification of Infant Cry Units (Cano-Ortiz, Escobedo-Becerro, 1999) (Cano, Escobedo and Coello, 1999). More recently, in (Orozco, & Reyes, 2003) we reported the classification of cry samples from deaf and normal babies with feed-forward neural networks. In 2004 Cano and his group, in (Cano, Escobedo, Ekkel, 2004) reported a radial basis network (RBN) to find out relevant aspects concerned with the presence of Central Nervous System (CNS) diseases. In (Suaste, Reyes, Diaz, and Reyes, 2004) we showed the implementation of a Fuzzy Relational Neural Network (FRNN) for Detecting Pathologies by Infant Cry Recognition.
Key Terms in this Chapter
Automatic Infant Cry Recognition (AICR): A process where the crying signal is automatically analyzed, to extract acoustical features looking to determine the infant’s physical state, the cause of crying or even detect pathologies in very early stages of life
Back propagation Algorithm: Learning algorithm of ANNs, based on minimising the error obtained from the comparison between the outputs that the network gives after the application of a set of network inputs and the outputs it should give (the desired outputs)
Learning Stage: A process to teach classifiers to distinguish between different pattern types.
Artificial Neural Networks: A network of many simple processors that imitates a biological neural network. The units are connected by unidirectional communication channels, which carry numeric data. Neural networks can be trained to find nonlinear relationships in data, and are used in applications such as robotics, speech recognition, signal processing or medical diagnosis
Hybrid Intelligent System: A software system which employs, in parallel, a combination of methods and techniques from Soft Computing
Fuzzy Relational Neural Network (FRNN): A hybrid classification model combining the advantages of fuzzy relations with artificial neural networks.
Fuzzy Sets: A generalization of ordinary sets by allowing a degree of membership for their elements. This theory was proposed by Lofti Zadeh in 1965. Fuzzy sets are the base of fuzzy logic.