Multimodal fusion of pinyin sequences and acoustic features

Multimodal fusion of pinyin sequences and acoustic features improve cognitive impairment detection accuracy in silico (accuracy 93.80%, AUC 0.93).

“The proposed method achieved an accuracy of 93.80% and an Area Under the Curve (AUC) of 0.93”

In silicoin Chinese speakersn = 300

A Multimodal Fusion Framework for Early Detection of Cognitive Impairment in Chinese Speakers Using Pinyin Sequences and Acoustic Features

AAIC 2025 abstract·Dec 24, 2025