Søren Riis

Followers

Following

Public Views

Interests

Uploads

Papers by Søren Riis

Development of an Australian behavioural method for assessing listening task difficulty at high speech intelligibility levels

International Journal of Audiology

OBJECTIVE To develop and validate an Australian version of a behavioural test for assessing liste... more OBJECTIVE To develop and validate an Australian version of a behavioural test for assessing listening task difficulty at high speech intelligibility levels. DESIGN In the SWIR-Aus test, listeners perform two tasks: identify the last word of each of seven sentences in a list and recall the identified words after each list. First, the test material was developed by creating seven-sentence lists with similar final-word features. Then, for the validation, participant's performance on the SWIR-Aus test was compared when a binary mask noise reduction algorithm was on and off. STUDY SAMPLE All participants in this study had normal hearing thresholds. Nine participants (23.8-56.0 years) participated in the characterisation of the speech material. Another thirteen participants (18.4-59.1 years) participated in a pilot test to determine the SNR to use at the validation stage. Finally, twenty-four new participants (20.0-56.9 years) participated in the validation of the test. RESULTS The results of the validation of the test showed that recall and identification scores were significantly better when the binary mask noise reduction algorithm was on compared to off. CONCLUSIONS The SWIR-Aus test was developed using Australian speech material and can be used for assessing task difficulty at high speech intelligibility levels.

Improving prediction of protein secondary structure using structured neural networks and multiple sequence alignments

Journal of computational biology : a journal of computational molecular cell biology, 1996

The prediction of protein secondary structure by use of carefully structured neural networks and ... more The prediction of protein secondary structure by use of carefully structured neural networks and multiple sequence alignments has been investigated. Separate networks are used for predicting the three secondary structures alpha-helix, beta-strand, and coil. The networks are designed using a priori knowledge of amino acid properties with respect to the secondary structure and the characteristic periodicity in alpha-helices. Since these single-structure networks all have less than 600 adjustable weights, overfitting is avoided. To obtain a three-state prediction of alpha-helix, beta-strand, or coil, ensembles of single-structure networks are combined with another neural network. This method gives an overall prediction accuracy of 66.3% when using 7-fold cross-validation on a database of 126 nonhomologous globular proteins. Applying the method to multiple sequence alignments of homologous proteins increases the prediction accuracy significantly to 71.3% with corresponding Matthew's...

Download

Prediction of Beta Sheets in Proteins

Most current methods for prediction of protein secondary structure use a small window of the prot... more Most current methods for prediction of protein secondary structure use a small window of the protein sequence to predict the structure of the central amino acid. We describe a new method for prediction of the non-local structure called β-sheet, which consists of two or more β-strands that are connected by hydrogen bonds. Since β-strands are often widely separated in the protein chain, a network with two windows is introduced. After training on a set of proteins the network predicts the sheets well, but there are many false positives. By using a global energy function the β-sheet prediction is combined with a local prediction of the three secondary structures α-helix, β-strand and coil. The energy function is minimized using simulated annealing to give a final prediction.

Download

Hidden Markov models and neural networks for speech recognition

Download

The scale illusion detection task: Objective assessment of binaural fusion in normal-hearing listeners

The normal auditory system can fuse sounds from both ears into a single sound object (binaural fu... more The normal auditory system can fuse sounds from both ears into a single sound object (binaural fusion). This ability can be assessed subjectively by asking whether listeners perceive one or two sounds or by the scale illusion percept described by Deutsch (1975). The aim of the current study is to develop an objective task to measure binaural fusion. Twelve normal-hearing participants had to detect one deviant note within a stream composed of a repeating melody while simultaneously being presented with another stream of randomized notes. The experiment included 3 conditions. First, in a monaural condition both streams were presented to the same ear. Then, in a binaural condition every second note from each of the two streams was presented to the other ear. Finally, in a binaural control condition, the timbre of all the notes presented to one ear was altered severely, to prevent binaural fusion. The expected result was a better detection of deviant notes for listeners that are able to...

Download

A modification of the scale illusion into a detection task for assessment of binaural streaming

The Journal of the Acoustical Society of America

Binaural streaming by frequency-proximity was investigated without subjective listener-feedback b... more Binaural streaming by frequency-proximity was investigated without subjective listener-feedback by modifying the scale illusion of Deutsch [J. Acoust. Soc. Am. 57, 1156-1160 (1975)] into a detectiontask. Nineteen listeners had to detect one deviation within a repeating melody stream, while simultaneously presented with a randomized distractor stream. Every second note in each stream was presented to the opposite ear, requiring binaural streaming to detect the deviant. Listeners performed well in this test but adding interaural delay or timbre-difference let the listeners group by lateralization instead. This confirms the grouping by frequency-proximity. The method could be used to investigate binaural streaming in hearing-impaired patients, where interaural percepts might differ.

Download

Hearing Device Manufacturers Call for Interoperability and Standardization of Internet and Audiology

American journal of audiology, 2016

This article describes the perspectives of hearing device manufacturers regarding the exciting de... more This article describes the perspectives of hearing device manufacturers regarding the exciting developments that the Internet makes possible. Specifically, it proposes to join forces toward interoperability and standardization of Internet and audiology. A summary of why such a collaborative effort is required is provided from historical and scientific perspectives. A roadmap toward interoperability and standardization is proposed. Information and communication technologies improve the flow of health care data and pave the way to better health care. However, hearing-related products, features, and services are notoriously heterogeneous and incompatible with other health care systems (no interoperability). Standardization is the process of developing and implementing technical standards (e.g., Noah hearing database). All parties involved in interoperability and standardization realize mutual gains by making mutually consistent decisions. De jure (officially endorsed) standards can be ...

Self-organizing letter code-book for text-to-phoneme neural network model

Method and system for speech recognition

Speech recognition method and system

Hidden Neural Networks

Http Dx Doi Org 10 1162 089976699300016764, Mar 13, 2006

Multilingual text-to-phoneme mapping

Multilingual speech recognition system using text derived recognition models

On text-based language identification for multilingual speech recognition systems

Adaptive transition bias for robust low complexity speech recognition

2001 IEEE International Conference on Acoustics, Speech, and Signal Processing. Proceedings (Cat. No.01CH37221), 2001

The basis for all methods described in this paper is the application of an adaptive transition bi... more The basis for all methods described in this paper is the application of an adaptive transition bias to the sequences of phoneme models that represent spoken utterances. This offers significantly improved accuracy in phoneme based speaker independent recognition, while adding very little overhead to the overall system complexity. The algorithms are tested using the low complexity hybrid recognizer denoted hidden neural networks (HNN) on US English and Japanese speaker independent name dialing tasks. Experimental results show that our approach provides a relative error rate reduction of up to 47% over the baseline system

Prediction of Beta Sheets in Proteins

Most current methods for prediction of protein secondary structure use a small window of the prot... more Most current methods for prediction of protein secondary structure use a small window of the protein sequence to predict the structure of the central amino acid. We describe a new method for prediction of the non-local structure called ,8-sheet, which consists of two or more ,8-strands that are connected by hydrogen bonds. Since,8strands are often widely separated in the protein chain , a network with two windows is introduced. After training on a set of proteins the network predicts the sheets well, but there are many false positives. By using a global energy function the ,8-sheet prediction is combined with a local prediction of the three secondary structures a-helix, ,8-strand and coil. The energy function is minimized using simulated annealing to give a final prediction.

Download

Joint Estimation of Parameters in Hidden Neural Networks

It has been proven by several authors that hybrids of Hidden Markov Models (HMM) and Neural Netwo... more It has been proven by several authors that hybrids of Hidden Markov Models (HMM) and Neural Networks (NN) yield good performance in speech recognition. However, in many of the current hybrids the HMM and neural networks are trained separately and only combined during decoding. In this paper we propose a new hybrid called Hidden Neural Networks (HNN) where all parameters are trained discriminatively at the same time by maximizing the probability of correct classification. The probability parameters in the HMM are replaced by neural network outputs, and instead of the local normalization of parameters used in standard HMMs the HNN is normalized globally. On the task of classifying TIMIT phonemes into five broad classes the new hybrid obtains a recognition accuracy of 83.7%, whereas a standard HMM obtains 76.1%. 1. INTRODUCTION It is well known that standard HMMs are based on a number of assumptions which limit their static classification abilities. First of all, it is usually assumed ...

On text-based language identification for multilingual speech recognition systems

The demand for multilingual speech recognition systems is growing rapidly. Automatic language ide... more The demand for multilingual speech recognition systems is growing rapidly. Automatic language identification is an integral part of multilingual systems that use dynamic vocabularies. Most state-of-the-art automatic language identification approaches identify the language based on probabilities of phoneme sequences extracted from the acoustic signal. Such methods can, however, not be applied to language identification based on text alone. This paper compares three text-based language identification methods aimed particularly at very short segments of text as encountered in, e.g., name dialling or command word control applications. The first method is based on artificial neural networks, the second on decision trees and the third on n-gram letter statistics. We conducted a series of experiments and the neural network approach is clearly better in terms of generalization performance and complexity.

Low complexity speaker independent command word recognition in car environments

2000 IEEE International Conference on Acoustics, Speech, and Signal Processing. Proceedings (Cat. No.00CH37100), 2000

In this paper we compare a standard HMM based recognizer to a highly parameter efficient hybrid d... more In this paper we compare a standard HMM based recognizer to a highly parameter efficient hybrid denoted Hidden Neural Networks (HNNs). The comparison was done on a speaker independent command word recognition task aimed at car hands-free applications. Monophone based HMM and HNN recognizers were initially trained on clean Wall Street Journal British English data. Evaluation of these baseline models on noisy car speech data indicated superior performance of the HMMs. After smoothing to the car environment, however, an HNN with 28k parameters provided a relative error rate reduction of 23-53% over HMMs containing 21k-168k parameters. Due to the low number of parameters in the HNNs, they have a real-time decoding complexity 2-4 times below that of comparable HMMs. The low memory and computational requirements of the HNN makes it particularly attractive for implementation on portable commercial hardware like mobile phones and personal digital assistants.

Download

Hidden Markov models and neural networks for speech recognition

The Hidden Markov Model (HMMs) is one of the most successful modeling approaches for acoustic eve... more The Hidden Markov Model (HMMs) is one of the most successful modeling approaches for acoustic events in speech recognition, and more recently it has proven useful for several problems in biological sequence analysis. Although the HMM is good at capturing the temporal nature of processes such as speech, it has a very limited capacity for recognizing complex patterns involving more than rst order dependencies in the observed data sequences. This is due to the rst order state process and the assumption of state conditional independence between observations. Arti cial Neural Networks (NNs) are almost the opposite: they cannot model dynamic, temporally extended phenomena very well, but are good at static classi cation and regression tasks. Combining the two frameworks in a sensible way can therefore lead to a more powerful model with better classi cation abilities. The overall aim of this work has been to develop a probabilistic hybrid of hidden Markov models and neural networks and to evaluate this model on a number of standard speech recognition tasks. This has resulted in a hybrid called a Hidden Neural Network (HNN), in which the HMM emission and transition probabilities are replaced by the outputs of state-speci c neural networks. The HNN framework is characterized by: Discriminative training: HMMs are commonly trained by the Maximum Likelihood (ML) criterion to model within-class data distributions. As opposed to this, the HNN is trained by the Conditional Maximum Likelihood (CML) criterion to discriminate between di erent classes. CML training is in this work implemented by a gradient descent algorithm in which the neural networks are updated by backpropagation of errors calculated by a modi ed version of the forward-backward algorithm for HMMs. Global normalization: A valid probabilistic interpretation of the HNN is ensured by normalizing the model globally at the sequence level during CML training. This is di erent from the local normalization of probabilities enforced at the state level in standard HMMs. Flexibility: The global normalization makes the HNN architecture very exible. Any combination of neural network estimated parameters and standard HMM parameters can be used. Furthermore, the global normalization of the HNN gives a large freedom in selecting the architecture and output functions of the neural networks. v vi Postscript les of this thesis and all the above listed papers can be downloaded from the WWW-server at the Section for Digital Signal Processing. 1 The papers relevant to the work described in this thesis are furthermore included in appendix C-F of this thesis. Acknowledgments At this point I would like to thank Ste en Duus Hansen and Anders Krogh for their supervision of my Ph.D. project. Especially I wish to express my gratitude to Anders Krogh for the guidance, encouragement and friendship that he managed to extend to me during our almost ve years of collaboration. Even during his stay at the Sanger Centre in Cambridge he managed to guide me through the project by always responding to my emails and telephone calls and by inviting me to visit him. Anders' scienti c integrity, great intuition, ambition and pleasant company has earned him my respect. Without his encouragement and optimistic faith in this work it might never have come to an end. The sta and Ph.D. students at the Section for Digital Signal Processing are thanked for creating a very pleasant research environment and for the many joyful moments at the o ce and during conference trips. Thanks also to Mogens Dyrdahl and everybody else involved in maintaining the excellent computing facilities which were crucial for carrying out my research. Center for Biological Sequence Analysis is also acknowledged for providing CPU-time which made some of the computationally intensive evaluations possible. Similarly, Peter Toft is thanked for learning me to master the force of Linux. I sincerely wish to express my gratitude to Steve Renals for inviting me to work at the Department of Computer Science, University of She eld from February to July 1997. It was a very pleasant and rewarding stay. The Ph.D. students and sta at the Department of Computer Science are acknowledged for their great hospitality and for creating a pleasant research atmosphere. I'm especially grateful to Gethin Williams for the many discussions on hybrid speech recognizers and for proofreading large parts of this thesis. I'm indebted to Gethin for his many valuable comments and suggestions to improve this manuscript. Morten With Pedersen and Kirsten Pedersen are also acknowledged for their comments and suggestions to this manuscript. Morten is furthermore thanked for the many fruitful discussions we've had and for his pleasant company at the o ce during the years. The speech group and in particular Christophe Ris at the Circuit Theory and Signal Processing Lab (TCTS), Facult e Polytechnique de Mons is acknowledged for providing data necessary to carry out the experiments presented in chapter 9 of this thesis. The Technical University of Denmark is acknowledged for allowing me the opportunity of doing this work. Otto M nsteds foundation and Valdemar Selmer Trane og Hustru Elisa Trane's foundation is acknowledged for nancial support to travel activities. Last but not least I thank my family and friends for their support, love and care during the Ph.D. study. A special heartfelt thanks goes to my wife and little daughter who helped me maintain my sanity during the study, as I felt myself drowning in ambitions. Without their support this work would not have been possible.

Download

Søren Riis

Uploads

Papers by Søren Riis

Log In