Marco Petroni

Classification of infant cry vocalizations using artificial neural networks (ANNs)

The analysis of infant cry vocalizations has been the focus of a number of efforts over the past thirty years. Since the infant cry is one of the only means that an infant has for communicating with its care-giving environment, it is thought that information regarding the state of an infant, such as hunger or pain, can be determined from an infants cry. To date, research groups have determined that adult listeners can differentiate between different types of cries auditorily, and at least one group has attempted to automate this classification process. This paper presents the results of another attempt at automating the discrimination process, this time using artificial neural networks (ANNs). The input data consists of successive frames of one of two parametric representations generated from the first second of a cry following the application of either an anger, fear, or pain stimulus. From tests conducted to date, it is determined that ANNs are a useful tool for cry classification and merit further study in this domain.

A comparison of neural network architectures for the classification of three types of infant cry vocalizations

The analysis of infant cry vocalization has been the focus of a number of efforts over the past thirty years. Since the infant cry is one of the only means that an infant has for communicating with its care-giving environment, it is thought that information regarding the state of an infant, such as hunger or pain, can be determined from cry vocalizations. To date, research groups have determined that a number of different types of cries can be determined auditorily and at least one group has attempted to automate this classification process. This paper presents the results of another attempt at automating the discrimination process, this time using artificial neural networks (ANNs). The input data consists of successive frames of 10 mel-cepstrum coefficients ranging in length from 0.75 seconds to 1 second. The mel-cepstrum coefficients were extracted from anger, fear, and pain cries. The ANNs used were a simple feed-forward network (FF), a recurrent neural network (RNN), and a time-delay neural network (TDNN). From tests conducted to date, it is determined that ANNs are a useful tool for cry classification and merit further study in this domain.

Identification of pain from infant cry vocalizations using artificial neural networks (ANNs)

The analysis of infant cry vocalizations has been the focus of a number of efforts over the past thirty years. Since the infant cry is one of the only means that an infant has for communicating with its care-giving environment, it is thought that information regarding the state of an infant, such as hunger or pain, can be determined from an infants cry. To date, research groups have determined that adult listeners can differentiate between different types of cries auditorialy, and at least one group has attempted to automate this classification process. This paper presents the results of another attempt at automating the discrimination process, this time using artificial neural networks (ANNs). The input data consists of successive frames of one or two parametric representations generated from the first second of a cry following the application of either an anger, fear, or pain stimulus. From tests conducted to date, it is determined that ANNs are a useful tool for cry classification and merit further study in this domain.

Database design of an intensive care unit patient data management system

A patient data management system (PDMS) developed for use in the intensive care unit (ICU) of the Montreal Childrens Hospital is described. The PDMS acquires real-time patient data from a network of physiological bedside monitors, and facilitates the review and interpretation of this data by presenting it as graphical trends, charts, and plots on a color video display. The data management structure integrates varied data types and provides database support for different applications, while preserving the real-time acquisition of network data. This structure is based primarily on OS/2 Extended Edition relational database. The relational database design is expected to solve the query shortcomings of the previous data management structure, as well as offer support for security and concurrency.<<ETX>>

A new, robust vocal fundamental frequency (F/sub 0/) determination method for the analysis of infant cries

The extraction and analysis of the fundamental frequency (F/sub 0/) of infant vocalizations has been the focus of a number of efforts over the past thirty years. It is thought that this parameter is an important information channel from which information regarding the state of an infant can be determined. To date, research groups working to extract the vocal fundamental frequency of infant cries have been limited in the resolution and granularity of the F/sub 0/ extraction methods used for adult speech, which, typically, are not well suited for infant cries. In general, the fundamental frequency range of adult speech is limited to values below 600 Hz, whereas for infant cries, F/sub 0/ can have a range of several hundred hertz, and be subject to rapid changes in certain cases. This paper presents a new method for accurately determining the F/sub 0/ of infant cries, which is applicable to other vocalizations as well. The method presented uses the crosscorrelation of adjacent speech segments to generate a three-dimensional plot called a crosscorrelogram. From this plot, the fundamental frequency of an infant cry can easily be extracted, regardless of the value of F/sub 0/.<<ETX>>

The design of a simulator for an intensive care unit patient data management system

A patient data management system (PDMS) for use in the intensive care unit (ICU) of the Montreal Childrens Hospital is described. The PDMS acquires real-time patient data from a network of physiological bedside monitors and facilitates the review and interpretation of this data by presenting it as graphical trends, charts, and plots on a color video display. A patient data simulator was developed for the purpose of complete testing of the PDMS. The simulator also allows for training the medical staff independently of hospital facilities and at a remote location from the ICU. This work outlines the design and functionality of the simulator and focuses on its implementation using OS/2s presentation manager window environment. Some experimental results and performance evaluations are presented, along with areas of future development and use.<<ETX>>

A patient data management system for an intensive care unit

A patient data management system (PDMS) based on a local area network linking fourteen bedside instrument monitors to a personal computer is presented. The PDMS acquires real-time data and graphically displays their trends using interactive menus. Fluid balance data, medical dosage calculations based on the patient data, and medical observations are also managed by the PDMS. It provides this information for online use by the medical personnel, produces the special forms and summary reports, and archives this data for storage and retrieval. The PDMS is designed and developed for an IBM PS/2 running under the OS/2 multitasking operating system and equipped with an 8514/A high-resolution color video display. The design of the PDMS is described here, with added emphasis on its software structure. Sample results are presented.<<ETX>>

Data modeling and design of a patient data management system in an intensive care unit

A patient data management system (PDMS) is being developed for use in intensive care units of the Montreal Childrens Hospital. The PDMS has a diversity of applications that handle various types of data. The numerous functionalities and management of different data types present the following design requirements: abstract definition of data types; unified view of data during the development phase; distinct levels of data management; and minimum coupling of the systems modules and higher degree of system flexibility. The authors outline the current database design and some aspects of its implementation using OS/2 Extended Edition. The multilevel data storage management is expected to solve the problem of storing the large volume of data and preserving the real-time response of the system.<<ETX>>

Real-time trend analysis for an intensive care unit patient data management system

A real-time trend analysis module design which is currently being developed for the patient data management system (PDMS) at the pediatric intensive care unit of the Montreal Childrens Hospital is discussed. The PDMS is based on a personal computer acquiring, in real time, patient data from a local area network of 14 bedside monitors, and displaying their trends graphically, among other tasks. The system is based on an IBM model 50 running under the OS/2 multitasking operating system and uses the 8514/A high resolution color video display. This work presents the design and implementation of the module based on two different supervised neural networks using the general delta rule learning mechanisms. The integration of such a module with a diagnosis expert system is also discussed.<<ETX>>

A robust and accurate cross-correlation-based fundamental frequency (F/sub 0/) determination method for the improved analysis of infant cries

The extraction and analysis of the fundamental frequency (F/sub 0/) of infant vocalizations has been the focus of a number of efforts over the past thirty years. It is thought that this parameter is an important information channel from which information regarding the state of an infant can be determined. To date, research groups working to extract the vocal fundamental frequency of infant cries have been limited in the resolution and granularity of the F/sub 0/ extraction methods used for adult speech, which, typically are not well suited for infant cries. In general, the fundamental frequency range of adult speech is limited to values below 600 Hz, whereas for infant cries, F/sub 0/ can have a range of several kilo hertz, and be subject to rapid changes in certain cases. This paper presents a new method for accurately determining the F/sub 0/ of infant cries, which should be applicable to other vocalizations as well. The method presented uses the crosscorrelation of adjacent speech segments to generate a three-dimensional plot called a crosscorrelogram. From this plot, the fundamental frequency of an infant cry can easily be extracted, with increased precision.