Abstract
The AKiD (Acoustic and Kinematic Characteristics of Speech in Dementia) project investigated the acoustic and articulatory characteristics of speech in individuals with Mild Cognitive Impairment (MCI) and dementia, with the long-term goal of identifying early speech markers that may support the diagnosis and monitoring of neurodegenerative diseases. AKiD addressed an important gap in current research. While previous studies have demonstrated that speech may provide valuable digital biomarkers of dementia, most investigations have focused on lexical, syntactic, semantic, or prosodic features. In contrast, the phonetic mechanisms underlying speech alterations in dementia have remained largely unexplored. The project therefore adopted an innovative multidisciplinary approach combining acoustic phonetics, articulatory phonetics, clinical linguistics, neurology, and statistical modelling to investigate how cognitive decline affects speech production.
Results achieved
AKiD successfully achieved all its scientific goals, making a significant contribution to research on clinical linguistics, experimental phonetics, and digital linguistic biomarkers for cognitive decline. One of the project's major achievements was the creation of the largest Italian speech corpus specifically designed for phonetic investigations of dementia. A total of 126 participants were recruited and carefully balanced across three groups: 42 individuals with Mild Cognitive Impairment or dementia, 42 healthy elderly controls, and 42 healthy adult controls. Overall, more than 43 hours of speech recordings were collected through both controlled reading tasks and semi-spontaneous speech production. In addition, a subgroup of participants underwent speech recording using Ultrasound Tongue Imaging (UTI), allowing the collection of articulatory data describing tongue movements during speech production. To date, no comparable Italian resource exists for the investigation of speech production in dementia. The project also produced standardized protocols for participant recruitment, speech elicitation, acoustic annotation, articulatory annotation, and statistical analysis. These protocols provide methodological resources that can be reused in future research and contribute to the standardization of experimental procedures in clinical phonetics. Acoustic recordings were transcribed and phonetically annotated using dedicated protocols developed within the project, while articulatory data were processed using state-of-the-art software for tongue contour extraction and movement analysis. The analyses carried out within AKiD produced important new evidence regarding the relationship between cognitive decline and speech production. Preliminary findings demonstrate that dementia affects several fine-grained phonetic characteristics of speech, including the stability and variability of vowel formants and other acoustic parameters. The project also provided the first articulatory evidence describing how tongue movements change during speech production in individuals with cognitive decline. These findings support the hypothesis that subtle phonetic and articulatory variations may represent promising digital biomarkers of neurodegenerative diseases and open new perspectives for future research on early diagnosis. Beyond its scientific findings, AKiD generated a substantial amount of high-quality research resources. FAIR principles guided the management and dissemination of project outputs whenever ethically possible. Several datasets derived from the corpus, annotation protocols, statistical scripts, and acoustic and articulatory resources have been made openly available through Zenodo using persistent identifiers (DOIs). Because voice recordings and ultrasound images constitute sensitive biometric data, raw recordings cannot be publicly released in accordance with ethical approvals and data protection regulations. Nevertheless, processed datasets and methodological resources have been shared to maximize transparency, reproducibility, and reuse by the scientific community. The project also produced a strong scientific dissemination output. Research findings were presented at numerous national and international conferences in the fields of clinical linguistics, phonetics, speech sciences, and communication disorders. Several scientific papers have also been published. The extensive corpus collected during the project will continue to support future publications addressing consonant production, prosody, rhythm, articulatory dynamics, longitudinal analyses, and automatic speech processing for dementia. Public engagement and knowledge transfer represented another important outcome of the project. Research results were disseminated through public lectures, outreach events, seminars, doctoral courses, international training activities, and workshops involving researchers, students, clinicians, and members of the general public. These activities contributed to increasing awareness of the potential role of speech analysis in supporting the early identification of neurodegenerative diseases while promoting interdisciplinary collaboration between linguistics, medicine, and engineering. Throughout its implementation, the project fully complied with the principles of Open Science, gender equality, equal opportunities, and responsible research established by the European Union. Open-access publication was prioritized whenever possible, FAIR data management principles were adopted for all shareable outputs, and all research activities were conducted in accordance with the relevant ethical approvals and data protection regulations. Overall, AKiD has established a unique research infrastructure for the study of speech production in dementia and has substantially advanced our understanding of the phonetic and articulatory correlates of cognitive decline. Although further validation studies are required before these findings can be translated into clinical screening tools, the project demonstrates that speech production contains measurable information that may contribute to future non-invasive, low-cost, and accessible approaches for the early detection of neurodegenerative diseases. Updated information on project activities, publications, datasets, and research outputs is available on the project website: https://site.unibo.it/akid/en.Project details
Unibo Team Leader: Gloria Gagliardi
Unibo involved Department/s:
Dipartimento di Filologia Classica e Italianistica
Coordinator:
ALMA MATER STUDIORUM - Università di Bologna(Italy)
Total Eu Contribution: Euro (EUR) 242.871,00
Total Unibo Contribution: Euro (EUR) 160.221,00
Project Duration in months: 24
Start Date:
30/11/2023
End Date:
28/02/2026
