Electrical inner ear prostheses like Cochlear Implants (CIs) help deaf and severely hearing-impaired persons to regain many of their communication abilities. The performance of CI in social environments is, however, not optimal. The new “Machine learning To Enhance teMPoral cOding foR cochleAr impLants” (TEMPORAL) project, in which Thomas Bäck and Anna Kononova from the Leiden Institute for Advanced Computer Science (LIACS) are involved, will examine how to improve their performance.
Fluctuating noise levels
‘At the moment, CI has a significant limitation in properly conveying detailed time information of sound to the auditory nerve,’ Kononova explains. ‘This means that in complex listening environments, like situations where the noise level is fluctuating, people with CI have difficulties understanding speech. Also, CI still has difficulties with things like music perception. Based on the computer modeling, artificial intelligence and patient-related research, the project will improve the speech coding strategies and tuning of the CI speech processor in order to improve those.’
‘The tuning will be done by a combination of machine learning algorithms and the so-called evolutionary algorithms, which are optimization methods gathered from the mode of evolution,’ Kononova explains. ‘The computational model of the stimulated cochlea will be combined with nonlinear optimization and machine learning algorithms to evaluate and develop a new and improved speech coding strategy. Furthermore, this will help to establish an efficient method for fine-tuning the CI processor, so that for various difficult auditory situations a realistic output is created.’
Improved hearing abilities
The researchers want to ensure that the models work well in real-life circumstances, which is why they pay special attention to the transition from the training environment to the real-world conditions. Kononova: ‘This will improve perception and lead to better communication and social participation for deaf and hard of hearing people, resulting in a better quality of life.’
TEMPORAL is a collaboration of the Leiden University Medical Centre (Johan Frijns, Jeroen Briaire), the Leiden Institute of Advanced Computer Science (Thomas Bäck, Anna Kononova) and Advanced Bionics (Hannover, Germany).