Speech, Hearing and Language: work in progress
Volume 12
2000
ISSN: 1470-8507
Abstracts


 Menu

Faulkner  Estebas  Ortega  Baker  Haydari

Simulation of the effects of cochlear implant electrode insertion depth for tonotopically-mapped speech processors
Andrew FAULKNER, Stuart ROSEN and Deborah STANTON

Abstract
It has been claimed that speech recognition with a cochlear implant is strongly dependent on the frequency alignment of analysis bands in the speech processor with characteristic frequencies (CFs) at electrode locations. However implanted electrode arrays often have the most apical electrode at positions with CFs of 1 kHz or more. The use of filters aligned in frequency with arrays in relatively basal locations inevitably leads to the loss of lower frequency speech information. This study simulates the effects on speech recognition of array insertion depth for a frequency-aligned speech processor in order to assess the significance of this information loss. Noise-excited vocoders were used to simulate a Continuous Interleaved Sampling (CIS) processor driving eight electrodes 2mm apart. Analysis filters always had centre frequencies matching the CFs of the simulated stimulation sites. The simulated position of the most apical stimulation site relative to the cochlear base was varied in 2mm steps between 24.9 mm (CF = 502 Hz) and 16.9 mm (CF = 1851 Hz). Identification of consonants, vowels and key words in sentences was measured in each condition. Each measure showed a significant decline in performance between each of the three more basal simulated electrode configurations. The results suggest that if implant processors were to use analysis filters frequency-aligned to electrode CFs, patients whose most apical electrode is 19 mm or less from the cochlear base are likely to suffer a significant loss of speech information.

 


 

 

 

Designed and built by
Martyn Holland
July 2000.
Click to comment.

Faulkner  Estebas  Ortega  Baker  Haydari  

Peak F0 downtrends in Central Catalan neutral declaratives
Eva ESTEBAS-VILAPLANA

Abstract

This paper examines peak fundamental frequency (F0) downtrends in Central Catalan sentences produced with a neutral declarative intonation. The tendency of an F0 decline over the course of an utterance is a well-known property of speech. However, the interpretation of such a downtrend varies in the literature. Whereas sometimes the downward trend of pitch has been analysed as a global effect, conceived as a component of the overall contour, at other times it has been treated as a local, phonologically controlled mechanism, which affects accents individually and in relation to previous accents. In order to determine the nature of F0 peak downtrends in Central Catalan, 192 sentences produced by four speakers were analysed. The results suggest that peak (or H(igh) accent) downtrends in Central Catalan are better explained as a linguistically controlled accent-by-accent downstep than as a global time-dependent declination. Thus, peak height in Central Catalan can be accurately predicted as a constant proportion of the height of the previous peak, except for the utterance-final H which undergoes a greater amount of lowering than that expected by the downstep rule. Final H values are better explained by means of a lowering constant, which is higher than the downstep ratio.


 

Faulkner  Estebas  Ortega  Baker  Haydari

Automatic cue-enhancement of natural speech for improved intelligibility
Marta ORTEGA, Valerie HAZAN and Mark HUCKVALE

Abstract

In previous work, 'cue-enhancement' was found to significantly increase the intelligibility of speech in noise. However, the practical application of the technique was limited by the fact that the regions of the speech signal to be enhanced needed to be manually labelled. The principal aim of this project was therefore to automate the identification and enhancement of 'landmark' regions containing a high density of acoustic cues and to demonstrate improvements in intelligibility at least equal to that obtained for manually-enhanced materials.

We have implemented a technique for automatic cue-enhancement via the automatic identification of potential enhancement regions (PERs), and evaluated intelligibility for automatically-enhanced speech, relative to natural or manually-enhanced speech. Little loss in intelligibility was seen between the manually-tagged and automatically-enhanced materials. However, there was little evidence of statistically-significant improvements as a result of the enhancements. This may have been due in part to the fact that amplification levels across consonantal regions had to be standardised, due to the limitations of the automatic tagging.


 

Faulkner  Estebas  Ortega  Baker  Haydari

No right ear advantage in gap detection
Richard J. BAKER, Stuart ROSEN and Azahara GODRICH

Abstract

It is well known that language function is based in the left-hemisphere in the vast majority of neurologically normal adults. There are a number of claims that this specialisation extends to, and is in fact based on, a specialisation for dynamic auditory nonspeech contrasts. Here we examine the extent of hemispheric asymmetry for a particular temporally-based task gap detection which has had conflicting reports about the extent to which it is better processed in one ear or the other in behavioural tasks. Three experiments were carried out in the same five normal hearing listeners, all involving the detection of short gaps in otherwise continuous flat-spectrum broadband noise presented via headphones. Stimulus presentation was monaural or dichotic (in which an identical noise burst without a gap was always presented contralaterally to the test ear). Various experiments used combinations of two-interval and one-interval tasks, adaptive and fixed level procedures, and blocked and random presentations. There was no effect of ear of presentation, nor dichotic vs. monaural presentation in any experiment. In short, we found no ear asymmetry in gap detection.


 

Faulkner  Estebas  Ortega  Baker  Haydari

Construction of a rotary vibrator and its application in human tactile communication
Abbas HAYDARI and Stuart ROSEN

Abstract

A major problem in the use of multichannel tactile communication devices is the extent to which multiple sites of stimulation interact and interfere with one another. In an attempt to minimise this interaction, we describe the design and construction of a novel rotary vibrator, based on the application of torsional oscillations to the skin. Preliminary psychophysical experiments are reported which compare the properties of rotary vibration to the more commonly used perpendicular vibration.


 
 


Other links:


link to UCL home page link to Phonetics and Linguistics home page