Raymond L. Watrous

Induction of finite-state languages using second-order recurrent networks

Second-order recurrent networks that recognize simple finite state languages over {0,1}* are induced from positive and negative examples. Using the complete gradient of the recurrent network and sufficient training examples to constrain the definition of the language to be induced, solutions are obtained that correctly recognize strings of arbitrary length.

The Impact of Computer-assisted Auscultation on Physician Referrals of Asymptomatic Patients with Heart Murmurs

As many as 50–70% of asymptomatic children referred for specialist evaluation or echocardiography because of a murmur have no heart disease.

Computer-aided auscultation of the heart: from anatomy and physiology to diagnostic decision support.

There is a clear and present need for computer-aided auscultation of the heart which arises from the highly informative nature of heart sounds, the inherent difficulty of auscultation and increasing pressure in healthcare for rapid, accurate, objective, documented and cost-effective patient evaluation and diagnostic decision making. There are advanced signal processing technologies that hold promise for developing computer-aided auscultation solutions that are intuitive, efficient, informative and accurate. Computer-aided auscultation offers an objective, quantitative and cost-effective tool for acquiring and analyzing heart sounds, providing archival records that support the patient evaluation and referral decision as well as serial comparisons for patient monitoring. There is the further promise of new quantitative acoustic measures and auscultatory findings that have more precise correlation with underlying physiological parameters. These solutions are being developed with the benefits of a rich literature of clinical studies in phonocardiography, the added insights derived from echocardiography, and advances in signal processing technology

Detection of the first and second heart sound using probabilistic models

In this study, a methodology based on hidden Markov models (HMM) as a probabilistic finite state-machine to model systolic and diastolic interval duration is proposed. The detection of the first (S1) and second (S2) heart sound is performed using a network of two HMMs with grammar constraints to parse sequence of systolic and diastolic intervals. Duration modeling was considered in the HMM model architecture selection based on experimental measurements of systolic and diastolic intervals in normal subjects. Feature extraction of heart sound signals was based on time-cepstral features. Results are presented in terms of detection performance compared with QRS peak annotations of the simultaneous ECG recording. The performance of the proposed approach has been evaluated in 80 subjects. The results showed that the system was effective to detect the first and second heart sounds with sensitivity of 95% and a positive predictive value of 97% and thus provides a promising methodology for heart sound analysis.

Speaker normalization and adaptation using second-order connectionist networks

A method for speaker normalization and adaption using connectionist networks is developed. A speaker-specific linear transformation of observations of the speech signal is computed using second-order network units. Classification is accomplished by a multilayer feedforward network that operates on the normalized speech data. The network is adapted for a new talker by modifying the transformation parameters while leaving the classifier fixed. This is accomplished by backpropagating classification error through the classifier to the second-order transformation units. This method was evaluated for the classification of ten vowels for 76 speakers using the first two formant values of the Peterson-Barney data. The results suggest that rapid speaker adaptation resulting in high classification accuracy can be accomplished by this method.

Connected recognition with a recurrent network

Abstract We attempted multi-talker, connected recognition of the spoken American English letter names b, d, e and v, using a recurrent neural network as the speech recognizer. Network training was based on forward-propagation of unit potentials, instead of back-propagation of unit errors in time. The target function was based on an input speech parameter which turns on and off at each onset of a spoken letter name. The network was trained to copy that input speech parameter to the output unit assigned to the correct letter name. Letter name discrimination was as high as 85% on test utterances.

Current status of Peterson-Barney vowel formant data

A question concerning the status of the Peterson-Barney vowel formant data is raised. Two machine-readable copies of the data were located, compared, and found to contain minor discrepancies. These discrepancies were resolved by comparison with a listing of the original data.

Phoneme discrimination using connectionist networks

The application of connectionist networks to speech recognition is assessed using a set of eight representative phonetic discrimination problems chosen with respect to a theory of phonetics. A connectionist network model called the temporal flow model (TFM) is defined which represents temporal relationships using delay links and permits general patterns of connectivity. It is argued that the model has properties appropriate for time varying signals such as speech. Networks are trained using gradient descent methods of iterative nonlinear optimization to reduce the mean‐squared error between the actual and the desired response of the output units. Separate network solutions are demonstrated for all eight phonetic discrimination problems for one male speaker. The network solutions are analyzed carefully and are shown in every case to make use of known acoustic phonetic cues. The network solutions vary in the degree to which they make use of context‐dependent cues to achieve phoneme recognition. The network ...

Context‐modulated discrimination of similar vowels using second‐order connectionist networks

Discrimination of medial adjacent vowels in the context of voiced and unvoiced stop consonants using connectionist networks is investigated. Separate discrimination networks were generated for one speaker from samples of the vowel centers of [e,ae] for the six contexts [b,d,g,p,t,k]. A single context‐independent network was similarly generated. The context‐specific error rate was 1%, whereas the context‐independent error was 9%. A method for merging isomorphic networks into a single network is described. The method uses singular value decomposition to find a minimal basis for the set of context‐specific weight vectors. Context‐dependent linear combinations of the basis vectors may then be computed using second‐order network units. Compact networks can thus be obtained in which the vowel discrimination surfaces are modulated by the phonetic context. In a preliminary experiment, as the number of basis vectors was reduced from 6 to 3, the error rate increased from 1% to 3%. Experiments with nonlinear optimiza...

A Computational Model of Cardiovascular Physiology and Heart Sound Generation

A computational model of the cardiovascular system is described which provides a framework for implementing and testing quantitative physiological models of heart sound generation. The lumped-parameter cardiovascular model can be solved for the hemodynamic variables on which the heart sound generation process is built. Parameters of the cardiovascular model can be adjusted to represent various normal and pathological conditions, and the acoustic consequences of those adjustments can be explored. The combined model of the physiology of cardiovascular circulation and heart sound generation has promise for application in teaching, training and algorithm development in computer-aided auscultation of the heart.