The centre was established in 2003 and 9 Ph.D. theses have since been completed. We currently have 8 running Ph.D. projects. Each student has his/her own project, but since we all do research within hearing, there are great opportunities to discuss ideas, approaches and results with other students and colleagues.
Besides being financially supported by the industry, there are also possibilities to collaborate with national and international industrial partners and hospitals. Hearing research is a very multidisciplinary field and we provide a large and high-quality international network to other research groups. We are involved in several formalized European networks and it is common for our students to spend some time abroad (typically in Europe or the U.S.A) as a part of their studies.
Current PhD projects
You may search for more projects and further descriptions in DTU's research database ORBIT - Online Research Database In Technology
Sebastien Santurette
Neural coding and perception of pitch in the normal and impaired human auditory system
Pitch is an essential attribute of hearing, which allows us to perceive melodies and chords in music, but also contributes to speech intelligibility and sound source segregation. Whether our auditory system uses neural mechanisms based on temporal information, spectral information, or both, for pitch extraction, remains a central debate in pitch perception research. This project aims to address this fundamental question.
In order to distinguish between possible theories, the project relates pitch perception outcomes to measures of basic auditory functions. Psychophysical experiments are performed in normal-hearing and hearing-impaired listeners, using stimuli with specific spectrotemporal properties. Such an approach may reveal how and at which level of the auditory system pitch is represented, and tell us which modeling approach to favor.
If auditory prostheses can transmit the pitch of sounds to hearing-impaired patients in a better way, this will improve their ability to segregate speech sources and enjoy music.
To be completed: 2011
Email:
Iris Arweiler
Processing of spatial sounds in the impaired auditory system
A common complaint from people with hearing impairment is difficulty with speech communication, particularly when more than one person is talking at the same time or when background noise is present. The problem often persists even if hearing aids are used. In these complex conditions, normal-hearing listeners take advantage of early reflections which increase the speech level when they are integrated with the direct sound.
The benefit from early reflections for monaural and binaural speech intelligibility was measured for normal-hearing and hearing-impaired listeners in a virtual auditory environment. For both groups of listeners, speech intelligibility improved with increased early reflection energy. Hearing aids should therefore not only amplify the direct sound, but also integrate spatial information from early reflections.
With a better understanding of the underlying mechanisms involved in early reflection processing we will be able to suggest advanced hearing-aid processing and fitting strategies.
To be completed: 2011
Email:
Claus Christiansen
Investigation of release from masking in normal-hearing and hearing- impaired listeners
People with normal hearing typically show a significant increase in speech recognition when a stationary background noise is replaced by a fluctuating noise. In contrast, people with impaired hearing often show only very little benefit from gaps in noise. It has been hypothesized that certain auditory functions are specifically linked to the ability to listen-in-the-gaps of a fluctuating noise and that these functions are degraded in hearing-impaired listeners.
This project investigates which auditory processing deficits are responsible for the reduced release from masking in hearing-impaired listeners. A model of auditory signal processing is used to investigate how different auditory functions, such as loss of sensitivity and reduced spectral and temporal resolution, affect the ability to understand speech in fluctuating noise.
Identifying the deficits in the impaired auditory system responsible for reduced speech recognition in background noise will help to improve compensation strategies in hearing-aid signal processing algorithms.
To be completed: 2012
Email:
Filip Munch Rønne
Modelling human auditory evoked brain responses to complex sounds
When various auditory stimuli are presented to human subjects, it is possible to record auditory evoked potentials (AEPs) on the surface of the scalp. AEPs can be used as an objective tool to assess hearing deficiency, for example in new-born screening. AEP generation mechanisms are relatively well understood for basic stimuli such as clicks and tones. However, it is poorly understood how more complex stimuli like speech are represented in AEPs.
The aim of the project is to develop a computational model for AEP generation, based on state-of-the-art knowledge of neural signal processing in the peripheral auditory system. The idea is to quantitatively predict a large set of existing literature data obtained with basic stimuli and to test the ability of the same model to account for speech evoked responses.
The prediction of AEP patterns will help understand how complex sounds like speech are represented neutrally at several stages along the human auditory pathway. The study will also provide a basis for investigating effects of hearing impairment on AEPs, which is important for clinical diagnostics.
To be completed: 2012
Email:
Remi Decorsiere
Speech intelligibility enhancement using modern envelope and phase manipulations
The purpose of the project is to understand the role of a speech signal's temporal amplitude modulations and spectral periodicities on speech intelligibility. Such variations, or spectro-temporal modulations, are typical of many environmental and natural sounds including speech. In hearing-impaired listeners, the ability to understand speech in noisy and reverberant acoustical environments is typically strongly reduced, mainly because of deficits in the temporal and spectral resolution in these listeners. One important aspect of this project is to construct a state-of-the-art sound manipulation tool, a so-called vocoder, which is used to accurately and systematically modify the spectro-temporal modulations of the signal. The modified signals are used as test stimuli in speech intelligibility experiments with normal hearing and hearing-impaired listeners.
Consonant identification and sentence intelligibility tests are performed and the analysis of the response patterns will clarify which temporal, spectral, or combined spectro-temporal modulations are crucial for speech intelligibility in the given tasks. Based on these results, spectro-temporal enhancement will be applied in a hearing-aid application, using a master hearing-aid signal processing platform. The results of this project are expected to be of major relevance for advanced compensation strategies in modern hearing instruments.
To be completed in 2013
Email:
Jasmina Catic
Distance perception in impaired and aided-impaired hearing
The present PhD project will be concerned with distance perception (in particular externalization) research and its implications for modern binaural hearing aid technologies. The main research goals are:
· to better understand the main auditory mechanisms, cues, and signal features that are involved in distance perception,
· to clarify the influence of hearing impairment on distance perception,
· to analyse the impact of modern (binaural/bilateral) hearing aids on distance perception in aided HI listeners, and
· to advice on methods for optimizing distance perception in aided HI listeners and thereby to minimize the occurrence of internalization.
To be completed in 2013
Email:
Søren Jørgensen
Predicting the intelligibility of processed noisy speech based on the signal-to-noise ratio in the modulation domain
One of the major challenges of current hearing research is to solve the "noise reduction paradox" that refers to the apparent mismatch between predicted and actual speech intelligibility following noise reduction signal processing. Current amplitude modulation-based speech intelligibility models are successful when predicting effects of linear distortions such as noise and reverberation, but fail to accurately predict the effects of non-linear signal processing and noise reduction. The current models consider the reduction in speech modulations as the critical physical characteristic related to speech intelligibility. However, recent investigations indicated that the ratio of speech to noise energy in the modulation domain might be the crucial physical characteristic underlying speech intelligibility. In order to systematically investigate the relation between the speech-to-noise modulation ratio and speech intelligibility, this project will: (i) develop a functional model that is inspired by auditory processing principles and includes a modulation signal-to-noise ratio based metric; (ii) investigate to what extend the model can account for speech intelligibility for normal-hearing listeners in various experimental conditions with processed and/or distorted speech; and (iii) analyze to what extend the reduced intelligibility typically observed in hearing-impaired listeners can be accounted for in the same modeling framework.
The results of this project are expected to contribute to a better understanding of the critical physical indicators of speech intelligibility and provide insights into which processes cause problems for the hearing impaired system. Furthermore, the results will be of substantial practical relevance since an accurate speech intelligibility model will make the development and testing of hearing aid algorithms much more efficient.
To be completed in 2013
Email:
Marton Marschall
Characterizing human auditory processing in reverberant environments with multiple sound sources
One of the main challenges of current auditory research is to solve the so-called “cocktail party problem” that reflects the substantial difficulty of most hearing-impaired listeners to understand speech in complex acoustic environments with multiple sound sources. At the same time, it has not yet been understood how normal-hearing listeners almost effortlessly decode the acoustic scene in such scenarios, showing a performance that is far beyond that obtained with any state-of-the-art technical system (e.g., automatic speech recognizers). In order to systematically investigate human auditory processing in challenging acoustic environments, this project will: (i) define and create a set of critical and representative acoustic scenarios with varying degree of room reverberation and background noise and with different numbers of target speakers and interfering sources; (ii) characterize the physical modifications of sound attributes (like temporal envelope and fine structure, pitch, interaural level and time differences) as a consequence of the transmission through the considered simulated rooms; (iii) investigate the ability of normal-hearing and hearing-impaired listeners to perceptually organize sound (e.g., to perceptually segregate a mixture of sound sources into the different components); and (iv) analyze and simulate effects of binaural hearing-aid signal processing (e.g. level-dependent and time-varying compression, binaural synchronization as well as directional filtering) on the above objective and perceptual outcome measures. The results of the project are expected to provide a better understanding of the auditory system’s processing strategies and will be of practical significance in hearing-aid design, where “smart” hearing aids might enhance stimulus features to optimize the input to the impaired ears.
To be completed in 2013
Email:
Completed PhD projects
Gilles Pigasse
Deriving cochlear delays in humans using otoacoustic emissions and auditory evoked potentials
In order to obtain an estimate of the cochlear delay that is closer to the normally functioning human cochlea, the present project investigates non-invasive methods in normal hearing adults. These methods include: otoacoustic emissions (OAEs), auditory brainstem responses (ABRs) and auditory steady-state responses (ASSRs). A comparison between the three methods was made across and within subjects, in order to highlight the impact of inter-subject variability on the cochlear delay estimates. The estimates of the cochlear delay obtained with OAEs, ABRs and ASSRs were in good agreement with previously reported studies. The comparison between OAE and ABR latency estimates was made over a broader range of frequencies (0.5-8 kHz), compared to previous studies. Below about 2 kHz the OAE delay is twice the cochlear delay, as if the travelling wave went back and forth in the cochlea, as predicted in current theories of OAE generation.
This relation, however, does not hold for higher frequencies, calling into question the physical relation between OAE and ABR delay estimates. The comparison between ABR and ASSR latency estimates demonstrated similar rates of latency decrease as a function of frequency. It was further concluded, in this thesis, that OAE measurements are the most appropriate to estimate cochlear delays, since they had the best repeatability and the shortest recording time. Preliminary results are also given for an experiment using stimuli designed to compensate for OAE delays. These were designed to try and reproduce the success of similar stimuli now used routinely to improve ABR signal-to-noise ratio.
Completed: 2008.
Peripheral auditory processing and speech reception in impaired hearing
People with impaired hearing often encounter great difficulty in understanding speech, particularly with background noise. The benefit from hearing aids varies among listeners. It has been hypothesized that part of the difficulty arises from changes in the perception of sounds well above hearing threshold such as reduced frequency selectivity and deficits in the processing of carrier-temporal-fine-structure at the output of the inner-ear filters.
The scope of this project was to investigate relations between frequency selectivity, temporal-fine-structure processing, and speech reception in listeners with impaired hearing. For this purpose, behavioral listening experiments as well as objective measurements of auditory evoked potentials were used.
The gained insights into auditory processing in listeners with impaired hearing may have implications for models of impaired auditory processing as well as compensation strategies in modern hearing instruments.
Completed: 2009
Eric Thompson
Characterizing binaural processing of amplitude-modulated sounds
Binaural hearing provides benefits over monaural hearing in the detection of auditory signals in complex acoustic environments and can reduce negative effects of reverberation on speech intelligibility. Current models of speech intelligibility in rooms, based on the transmission of amplitude modulations in speech signals, do not consider the benefits of binaural listening and may underestimate speech intelligibility as a result.
Psychoacoustic experiments were performed to measure human abilities in the detection of sound amplitude fluctuations in anechoic and reverberant spaces, and to measure interactions between spatial hearing and amplitude modulation processing. These data and data from a consonant identification in reverberation experiments will be used to improve binaural models for signal detection and speech intelligibility.
This research will provide insight into human performance in complex listening environments, which will be used to improve models of hearing and assistive listening devices.
Tobias Piechowiak
Spectro-temporal analysis of complex sounds in the human auditory system
Many natural sounds including speech contain amplitude modulations. It is commonly assumed that a comparison of amplitude modulations across frequency by the auditory system plays an important role for the extraction of signals embedded in noisy backgrounds.
The objective of the project was to quantitatively simulate the processing of amplitude modulated sounds in the auditory system. In particular, we tested different signal processing mechanisms for signal enhancement in noisy backgrounds that utilize the coherence of signal information across frequency.
The modeling results from this project will be useful for optimizing the signal enhancement processing strategies in hearing instrumentation such as digital hearing aids.
Completed: September 2009
Jens Bo Nielsen
Assessment of speech intelligibility in background noise and reverberation
Reliable methods for assessing speech intelligibility are essential within hearing research, audiology, and related areas. Such methods can be used for obtaining a better understanding of how speech intelligibility is affected by various environmental factors or by different types of hearing impairment. Speech intelligibility tests are also essential for the development and fitting of hearing aids.
In this project, two sentence-based tests for speech intelligibility in Danish were developed; (i) the Conversational Language Understanding Evaluation (CLUE) and (ii) a modified version, which complies with an international standard intelligibility test. The validation of both tests showed that they produce reliable speech intelligibility assessments. The project also included an investigation of the influence of reverberation on speech intelligibility.
The two developed sentence tests are expected to be useful for assessing speech intelligibility with Danish normal-hearing and hearing impaired listeners.
Completed: December 2009
Helen Connor Sørensen
Hearing aid amplification at low input levels
People with a hearing loss often have a loss of audibility of soft sounds, such as distant voices and birds. This lack of audibility can be alleviated by means of a hearing aid with compression. Previous research from other research groups has produced conflicting results about how this compression should be fitted to the individual hearing loss for maximum benefit of the hearing aid user.
This project considered the hearing aid wearer preference for the hearing aid compression threshold and implicitly the gain at low input levels. Results have been collected using both laboratory listening experiments and field experiments. Results indicate that the preference for hearing aid compression threshold depends on the speed at which the compressor operates, as well as the listening environment and the hearing aid experience of the user.
Improving the hearing aid amplification for low input levels enhances the audibility for speech and environmental sounds, and it also improves the sound quality.
Completed: December 2009
Morten Løve Jepsen
Modelling auditory processing and speech perception in hearing-impaired listeners
Models of auditory processing and speech perception can help us to understand our auditory system. The consequences of different types of hearing losses for speech perception are not fully understood and there are enormous individual differences in performance across listeners, particularly when background noise is present.
The objective of this study is to develop an auditory processing model in order to simulate perceptual consequences of hearing impairment. First a model was designed to predict masking and discrimination data. Then this model was used to derive internal (auditory) representations of speech signals which were used to predict the performance in a consonant identification task in individual hearing-impaired listeners.
In hearing-aid development, the processing of the algorithms has been evaluated in time-consuming subjective listening tests. On the basis of this project, an auditory-model based system could be developed which objectively evaluates the effects of hearing-aid signal processing.
Completed: 2010
Sarah Verhulst
Characterising temporal nonlinear processes in the human inner ear using otoacoustic emissions
Otoacoustic emissions (OAEs) are faint echoes that arise in the inner ear as by-products of the hearing process and can be recorded from the human ear canal. They contain information about the functioning of the nonlinear gain mechanisms present in the hearing process, and are vital for the sharp tuning of the human auditory filters.
Temporal features of the inner ear gain mechanisms were investigated by the use of click-evoked otoacoustic emissions (CEOAEs). The experimental results were compared to simulations that were made with a nonlinear adaptive transmission-line model of the inner ear. It was found that temporal changes (0 - 10 ms) in the level of the CEOAE were probably due to temporal overlap of inner-ear impulse responses within this time frame.
Temporal changes in nonlinear gain mechanisms may have consequences for the onset detection of sound stimuli and provide an application for the design of compression characteristics in hearing aids.
Completed: 2010
Sylvain Favrot
Implementation and evaluation of a loudspeaker-based room auralization system
Humans are able to communicate in reverberant environments in the presence of disturbing talkers and noise. This ability is investigated in order to design and optimize modern speech and audio technologies, such as hearing aids, cochlear implants, perceptual audio-coders, or automatic speech recognition systems. These investigations require the reproduction of highly realistic acoustic scenarios.
This project aims at implementing and evaluating a loudspeaker-based room auralization system that generates a highly realistic environment suitable for investigating human auditory perception. This system combines acoustic room models with loudspeaker-based auralization techniques. Different multi-channel playback algorithms are implemented and their performance in various psychoacoustic experiments is assessed.
This platform will be useful for studying auditory processing and perception in normal-hearing and hearing-impaired listeners as well as testing hearing-aids in fully controlled and realistic environments.
Completed: 2010