[Application of artificial neural network in assessing speech intelligibility in adult patients after cochlear implantation].

OBJECTIVE

An objective of the study was to evaluate the feasibility of using Voice Extraction (VE) technology in adult patients after cochlear implantation and to conduct a comparative analysis of speech intelligibility of a noise-free signal and a signal using VE technology in adult patients after cochlear implantation.

MATERIAL AND

METHODS

Sixty-seven patients took part in the study. All patients were diagnosed with chronic bilateral severe to profound sensorineural hearing loss. All patients underwent unilateral cochlear implantation. The observation period for patients ranged from 6 months to 15 years. All patients underwent pure-tone free field audiometry using an optimally fitted speech processor of the cochlear implant. All patients underwent free field speech audiometry. Speech intelligibility was assessed in both silence and when speech material was presented against the background noise interference with street noise. Next, patients were presented with verbal articulation tables, processed through a voice extraction device. The voice extraction device is based on a trained neural network and operates autonomously. At the end of the study, the percentage of words correctly reproduced was calculated. Speech intelligibility improvement rates (%) were calculated equal to the difference between intelligibility after VE and intelligibility before VE.

RESULTS

For 55 study participants (82%), VE technology increased intelligibility by an average of 11.9% in absolute terms.

CONCLUSION

The application of Voice Extraction technology in the development of new options for speech coding by a speech processor, and as a result, in the rehabilitation of patients after cochlear implantation, seems to be a promising and effective method.