Seminar on Sound Quality

Half day seminar on models and measurement of Sound Quality

Time and place

October 11, 2010, 12:30-17:00
Widex A/S, Nymøllevej 6, 3450 Lynge, Denmark

The seminar is free of charge, however, limited to 110 participants.

Directions: Enter at the main entrance at 210 degrees
Arriving by car: Please park at the main parking lot and not in front of the main entry
Arriving by public transportation: From Allerød Station use bus 381 or 335. From Farum Station use bus 335.
Map: Google Map

Scope

The ability to quantify sound quality is obviously imperative for signal processing engineers working in the audio field. In some applications, sound quality is the primary design goal: process the audio signal to provide the listener with a more pleasant sound. In other situations sound quality is not the main design objective, e.g., when the purpose of the processing is to enhance or suppress certain characteristics in a signal, compress it, transmit it, etc. In such applications, sound quality still plays a prominent role and in general one will try to meet the primary design criteria in such a way that processing artifacts are as small as possible.

Often one will resort to rely on one's own judgment, as listening tests are costly and time-consuming while in a development phase. Being able to take a more formal approach and use reliable models for assessing sound quality is highly desirable. In this seminar, a broad introduction to the topic "sound quality" will be given and subsequently models and measurements of speech and audio quality will be presented.

The seminar is intended to be relevant for engineers working with audio processing in such fields as mobile phones, headsets, VOIP and hearing aids.

Program

 

12:30

Registration

12:45-12:55

Welcome by Magnus Nørgaard, Widex A/S, Denmark

12:55-13.50

Sensory evaluation of sound systemsNick Zacharov

Head of Department Dr. Nick Zacharov, SenseLab, DELTA, Denmark

Abstract:

This presentation will start with short overview of sound quality and its role in different applications (white good, hifi, telecommunications, etc.) and how this can be shaped by signal processing technologies. Sensory Evaluation techniques as a whole will be presented as one of the tool sets available for exploring and understanding the characteristics of sound quality perception both from the Expert and Consumer perspectives. A number of example cases shall be presented to illustrate methods (lab based and online testing) and they type of information that can be gleened from such analysis.

13:50-14:00

Coffee break

14:00-14.55

Measuring the impact if signal enhancement on the quality of noisy speech

Senior Lecturer Dr. Mark Huckvale, Centre for Law-Enforcement Audio Research, Department of Speech, Hearing and Phonetic Sciences, University College London, United Kingdom

Abstract: The CLEAR project researches into single-channel speech signal enhancement for law-enforcement applications. As well as investigating new methods for enhancement of noisy and distorted speech signals, we are also concerned with the evaluation of existing technologies. In this talk I will describe our work in evaluating the impact of noise-reduction on the ability of people to perform speech communication tasks.
Although it may be possible in principle for noise reduction methods to improve the intelligibility of speech corrupted with broad-band noise, in general we find that such methods have either no effect or a detrimental effect. Interestingly, despite this, users seem to prefer noise-reduced speech over unprocessed noisy speech in subjective preference tests. It is possible that noise-reduced speech is preferred because it is "easier to process" or is "less tiring to listen to", which raises the possibility that noise reduction might provide beneficial to users exposed to noisy speech throughout the working day. Our research investigates whether this is actually the case.

In recent experiments we have looked at how noise reduction changes the cognitive effort required to process noisy speech materials for which word intelligibility is high. I will report on the results of two experiments, one using a reaction-time task, and one using an audio proof-reading task. Such performance-based measures of speech signal quality may complement existing measures of intelligibility and quality in audio research.

14:55-15:25

Coffee break and networking

15:25-16:20

Application of auditory models to assessing the sound quality of hearnig aid signal processing algorithms

PD Dr. Volker Hohmann, Medical Physics Section, Carl von Ossietsky-Universität, Oldenburg, Germnay

Abstract: Advances in systems technology allow for increasingly complex processing algorithms in hearing systems addressing increasingly complex acoustic conditions. These developments have the potential of improving the rehabilitation of hearing impairment, but establishing reliable measures of benefit is quite difficult for these complex algorithms and conditions. Being the “gold standard” for algorithm evaluation, subjective testing of hearing-impaired subjects has some limitations in this context. It is very time-consuming and the long acclimatization time needed when listening with new devices in complex acoustic conditions cannot easily be accounted for. Therefore, objective methods for estimating the benefit of an algorithm in a certain acoustic condition are desirable. They allow for identifying promising candidate algorithms and the acoustic conditions in which the algorithms might be applicable and thus identify critical acoustic conditions to be tested subjectively. In addition and in combination with technical measures like segmental signal-to-noise ratio and distortion measures, perceptual measures based on auditory models might be useful for developing meaningful objective measures. This contribution presents recent applications of auditory models to objective algorithm evaluation, including results from a project on hearing aid sound quality prediction funded by the major hearing aid companies (CTO project). Because of the growing importance of binaural and multi-microphone processing in hearing instruments, novel approaches to binaural/multichannel models for speech intelligibility and quality will also be emphasized.

16:20-16:25

Break

16:20-16:55

Validation of objective sound quality models for hearing aids

Audiological Concept Developer, Dr. Lars Bramsløw, Oticon A/S, Denmark

Abstract: This study investigated three objective sound quality models developed for hearing aid test with hearing-impaired listeners. After design and training, the models were validated on a range of new signals and new types of signal processing using both normal-hearing and hearing-impaired listeners. The results of the validation were mixed, with some predictions highly correlated with subjective ratings and others poorly correlated. On average, the models seemed to perform quite equal, although specific models performed better on specific data sets.

16:55-17:00

Wrap up

 

The seminar is arranged in cooperation with:

 

Danish Sound Technology Network

Signal Processing Chapter of the IEEE Denmark Section

 

 

Member Comments

Kasper Eskelund Sunday, 10.10.2010 22:00
Ufortunately, I am unable to attend the seminar. Uregister button is not visible anymore - please unregister me.
Jan Larsen Friday, 08.10.2010 13:06
Registration is closed as the maxmimum limit for the lecture hall has been reached.
Renskje Hietkamp Friday, 08.10.2010 13:06
Ufortunately, I won\'t be able to attend. Uregister button is not visible anymore; please unregister me.
Jan Larsen Friday, 08.10.2010 13:06
Registration is closed as the maxmimum limit for the lecture hall has been reached.
Jan Larsen Friday, 08.10.2010 13:05
Registration is closed as the maxmimum limit for the lecture hall has been reached.
Page 1 ForwardPage 2 Forward

Next > Forward

Displaying results 1 to 5 out of 8

 Login