PathoVoiceAI : Classifying Pathology Types in Human Voices

초록

This paper proposes PathoVoiceAI, a technique that classifies the most prevalent pathologies in human voices. The proposed method employs Electroglottography (EGG) that reflects the changes in the vocal cord vibrations along with the voice data for enhanced accuracy in multiclass pathology classification. The preliminary results show that the use of multimodal inputs in the proposed method improves the classification accuracy by 9% in comparison to the conventional methods relying on single modality inputs. This highlights the effect of incorporating multimodal analysis in enhancing the correctness of diagnostic outcomes.