1. Fondazione Istituto Italiano di Tecnologia, Genoa, Italy
EVALUATION OF A COMPLEMENTARY HEARING AID FOR SPATIAL SOUND SEGREGATION
Luca Giuliani1, Luca Brayda1, Sara Sansalone2, Stefania Repetto2 and Michele Ricchetti2
Fondazione Istituto Italiano di Tecnologia, Genoa, Italy
2 Linear s.r.l, Genoa, Italy 1

2. Outline The cocktail party problem Beamforming State of the art
The Glassense system
Experimental setup and protocol
Data analysis and results
Conclusion and future works

3. The cocktail party problem
Firstly defined by Colin Cherry in 1953, consists in the trouble of perceiving speech in noisy social contests.
Colin Cherry

4. The cocktail party problem
Approximately one-third of people over 65 year are affected by disabling hearing loss.
People suffering from hearing impairment have bigger troubles in separating acoustical sources.
Limited hearing aids performance in noise.

5. Beamforming
Also called spatial filtering, is a signal processing technique used in sensor arrays to improve the Signal to Noise Ratio (SNR) of a target source.

6. State of the art
Modern behind-the-ear earing aids are equipped with more then one microphone to provide directionality.
The improvement deeply depends on the distance between the microphones and on their number.

7. State of the art
Several solutions has been proposed in the last years, but none of them made a real breakthrough in the market yet.
High directivity
Complementary to hearing aid
No need for calibration on hearing loss
Cannot use head as spatial selector
Not available on the market anymore
Hearing-aid necklace by Widrow (2001)

8. State of the art
Several solutions has been proposed in the last years, but none of them made a real breakthrough in the market yet.
High directivity
Use head as spatial selector
Substitutes hearing aid
Need for calibration on personal hearing loss
Hearing glasses by Varibel company (available on the market since 2008)

9. The Glassense Signal sent to hearing aids wireless
Spectacles with two arrays of digital microphones embedded in the temples.
Signal sent to hearing aids wireless

10. The Glassense
Each temple embeds a 10 cm array of four digital microphones.

11. The Glassense
Beam power pattern at various frequencies. The sounds coming from the sides and the rear are attenuated.

12. Is it possible to overcome spatial filtering limitations of traditional hearing aids to improve speech comprehension capabilities for hearing impaired people in noisy environments?

13. Subjects 8 subjects 6 hearing impaired 2 healthy hearing
Age years

14. Experimental Setup Anechoic room
Four speakers around the subject, 1m away at 0° (directly ahead), 180° (back) and ±90°
All subjects wore a couple of in-ear hearing aids

15. Experimental Setup Target signal:
Lists of 10 meaningful bisyllabic unrelated words (200 in total)
Reproduced by the frontal speaker

16. Experimental Setup Competing noise:
Four channel registration of a cocktail-like typical acoustic background
Reproduced by all the surrounding speakers (included the frontal one)

17. Experimental Protocol
Trial = listen and repeat task on target words with competing noise Performance = % of correctly repeated words Target signal calibrated on the subject to reach 100% performance without noise

18. Experimental Variables
Independent variable: SNR Level
Target speech volume fixed
Competing noise volume changed in each trial, i.e. different Signal to Noise Ratios (SNRs)
Dependent variable: Performance

19. Experimental Conditions
Hearing aids condition
sounds perceived directly by hearing aids microphones.

20. Experimental Conditions
Unfiltered condition
Glassense connected to hearing aids
sounds perceived by Glassense microphones, without spatial filtering

21. Experimental Conditions
Beamforming condition
Glassense connected to hearing aids
sounds perceived by Glassense microphones, with spatial filtering

22. Data Analysis Speech Reception Threshold (SRT)
SNR at which the subject correctly understand 50% of target words
Obtained interpolating subjects performances

23. Data analysis and results
Mean SRT comparison
improvement of 3.3 dB in beamforming condition with respect to unfiltered
improvement of 2.4 dB in beamforming condition with respect to hearing aids

24. Conclusion
The use of Glassense can improve the performance of traditional hearing aids in noise.

25. Conclusion
No need for calibration on hearing loss, since hearing aid is already calibrated.
Easy to swap between traditional hearing aid and Glassense.

26. Future works Extension of experimental population
Measurement of Glassense contribution with more hearing aid models
Comparison with other microphone array devices

27. Thank you! Co-funding: Claudio Lorini Davide Dellepiane Elio Massa
Francesco Diotalevi
Senior Technician
Luca Brayda
Team Leader
Linear s.r.l.
Michele Ricchetti
Project Coordinator
Sara Sansalone
Stefania Repetto
Petra Bianchi

28. Questions?

29. EVALUATION OF A COMPLEMENTARY HEARING AID FOR SPATIAL SOUND SEGREGATION
DEMO AVAILABLE!
/IITGlassense
@IITGlassense 1

30. IIT
Post-doc position Antropometric modeling for personalized sensory aids
PhD position Binaural filtering techniques for augmented perception and understanding of sounds
to: