The above video is a sample of in-clinic digital voice recording for Framingham Heart Study, performed by a research assistant. Using these voice recordings of participant responses to neuropsychological tests is an effective way to detect changes in cognition. Speaking is a cognitively complex task and embedded in the spoken responses are acoustic and linguistic features mapping onto cognitive domains implicated by neuropathological changes. Shifting cognitive capabilities are expressed through vocal responses in subtle ways.
Analysis of this data is carried out by a number of tools that are available on a public GitHub repository. It contains links to repositories that contain example workflows, READMEs, sample data, and Docker files that facilitate the usage of various open-source voice feature extraction packages, tools, datasets, and models developed by the FHS-BAP Data Core. These toolkits were developed to support scientific research surrounding investigations of relationships between brain aging and voice features, although the extraction of voice features does have wider applicability.
Examples of some of the information on our GitHub repository include tools to standardize digital voice audio files with varying metadata to a standard format, an example of utilizing openSMILE (an open-source toolkit for audio analysis) to generate acoustic features, and tools to evaluate language identification via various models on open-source datasets.
The following recent publications are based on FHS digital voice data:
Related Publications |
---|
Young CB, Smith V, Karjadi C, Grogan SM, Ang TFA, Insel PS, Henderson VW, Sumner M, Poston KL, Au R, Mormino EC. Speech patterns during memory recall relates to early tau burden across adulthood. Alzheimers Dement. 2024 Apr;20(4):2552-2563. doi: 10.1002/alz.13731. Epub 2024 Feb 13. PMID: 38348772; PMCID: PMC11032578. |
Ding H, Hamel AP, Karjadi C, Ang TFA, Lu S, Thomas RJ, Au R, Lin H. Association Between Acoustic Features and Brain Volumes: the Framingham Heart Study. Front Dement. 2023;2:1214940. doi: 10.3389/frdem.2023.1214940. Epub 2023 Nov 23. PMID: 38911669; PMCID: PMC11192548. |
Ding H, Mandapati A, Karjadi C, Ang TFA, Lu S, Miao X, Glass J, Au R, Lin H. Association Between Acoustic Features and Neuropsychological Test Performance in the Framingham Heart Study: Observational Study. J Med Internet Res. 2022 Dec 22;24(12):e42886. doi: 10.2196/42886. PMID: 36548029; PMCID: PMC9816957. |
To implement audio recording in research protocols, please see the following four sections:
-
Informed consent It is crucial to provide an in-depth description of the audio recording implementation and use of the data as part of the informed consent process. In addition, it is important to provide detailed information regarding data usage, storage, and protection. A copy of the FHS-BAP consent form can be found under Appendix B: FHS-BAP Consent Form.
-
Selecting Audio Recording Equipment For audio recording equipment, consider the following factors: portability, data storage space, battery life, ease of set up, sampling rate, etc. FHS-BAP uses the Zoom H4N recorder.
-
Storage and retrieval of audio recordings Use a standardized naming convention for the audio recording files and maintain a data log for inventory management.
-
Post data collection processing Generating analysis-ready files is resource intensive. It may not be possible for all researchers, especially if funding may be an issue. Personally identifiable information is a common concern for audio recordings; hence these files must undergo post data-collection processing before they can be made available for analysis, especially in situations where the sensitive data are being shared with other researchers.
This document is adapted from the manual published by the ADRC Clinical Task Force Cognitive Working Group, in collaboration between the Framingham Heart Study Brain Aging Program at Boston University and the Indiana Alzheimerβs Disease Research Center. It serves as a guide for researchers interested in audio recording of cognitive testing for research analysis.
Digital Voice Capture In-Person Recording Manualβ
Links to external digital voice resources:
- HuggingFace Audio Course: Introduction to audio data
- Google Cloud Audio Concepts: