Intelligible Car for Hearing Impaired – AIDA

Signal modifications aiming at improving speech intelligibility will be developed. Their efficiency will be evaluated on real listeners. As it is difficult to recruite subjects suffering from a given hearing loss, a simulator will be developed. This simulator is expected to modify sound so that a normal-hearing person can hear what a hearing-impaired one experiences.
Also, a man-machine interface will be built so that an user can adjust the parameters of the speech modification process according to his hearing profile.

Signal processing has been developed and implemented, as well as the hearing loss simulator. This simulator will allow to conduct listening experiments in the next few weeks. These experiments will evaluate the speech improvement, depending on the various loss profiles.

to be done

International conferences :
1. N. Grimault, A. Corneyllie, S. Garcia, E. Parizet, L. Brocolini. Real-time hearing impairements simulator for speech intelligibility measurements. SpiN (Speech in Noise workshop, 7-8 January 2016, Groningen)
2. K. Nathwani, M. Daniel, G. Richard, B. David, V. Roussarie. Formant shifting for speech intelligibility improvement in car noise environment. ICASSP 2016 (20-25 March 2016, Shangai)

National conference :
1. E. Parizet, N. Grimault, S. Garcia, A. Corneyllie, L. Brocolini. Simulateur de pertes auditives. 13ème Congrès Français d'Acoustique (11 – 15 Avril 2016, Le Mans)

In Europe, the estimated population suffering from hearing impairment is as large as 16%. Hearing impairment may be related to their age (presbyacousis), to excessive exposure to loud sounds (this is true for more and more young people) or to other reasons. In France, hearing aids are not widely used, as they are negatively perceived to be related to some disability (on the contrary to eye glasses) and remain too expensive. The individual fitting of these hearing aids is difficult and, quite often, people do not use them as they find them uncomfortable and not as efficient as expected. In the meantime, sound interactions increase in cars, and many of these sounds involve speech communication. Such interactions can be related to safety (driving aid device), functionality (GPS system) or comfort (telephone or radio). In the case of hand-free phone, many people experience a poor intelligibility and the attentional load required to focus on conversation might become dangerous. For such hearing impaired people, the only way they can try to improve the ease of hearing is to increase sound level, which is uncomfortable and non optimal, given their hearing abilities.

This project is dedicated to provide new tools to improve speech messages intelligibility in a noisy environment, by giving to the listener an easy-to-use device which would allow him to find the optimum set-up by his own. This user layer will hide sophisticated speech signal processings that will be adapted in real time according to the noise situation and the speech signal to improve.

The targeted application field will be the car industry, for three reasons :
- car buyers are significantly older (50 years old) than the average population;
- speech sounds are more and more used in a car for communication purpose;
- noise is constantly varying in the cabin.

To achieve this goal, the consortium is made of four partners :
- Telecom ParisTech will develop signal processing algorithms in order to improve speech intelligibility in noise. Some algorithms will reinforce the speech without modifying any important speaker feature, some other ones will use more important modifications (as an example, by slightly modifying formant places).
- PSA Peugeot Citroën will use MAX/MSP real time platform to develop man-machine interfaces. These interfaces will be used for subjective experiments and for conceiving prototypes at the end of the project.
- INSA-Lyon (Laboratoire Vibrations Acoustique) will evaluate the efficiency of algorithms and interfaces, as regard to speech intelligibility and quality.
- The Lyon Neuroscience Research Center will develop a system allowing to simulate various hearing impairments in real time. This system will be useful in the following to conduct easy-to-use experiments, without selecting participants with a targeted hearing impairment.
Also, it is worth to be mentioned that an association of audioprothesists (Association Française des Audioprothésistes Indépendants) will be associated to the project, as a sub-contractor of one partner. The input from this association will be a strong knowledge of hearing impairement and fitting procedures. This will be useful to the consortium within all the project (developing the device and testing it at the end on a population of hearing impaired listeners). Moreover, recycling the choices the patient made when using the device, could help the audioprostesist for the set-up of hearing aids parameters.
The project is structured by six tasks. One for each of the four above-cited partners and two additional ones: one will be devoted to evaluate the importance of each kind of hearing impairment in the population and to select some targeted population and the second is the managing of the overall project.