Acoustic assessment of speech in Frontotemporal Lobar Degeneration: Robust phonetic measures and tasks for its automatization

AcousA is a 3-year project funded by the Spanish Ministry of Science that aims to characterize motor speech in fronto-temporal lobar degeneration (FTLD) using phonetic variables and automatise its analysis (PID2022-139004OA-I00).

Show details


Frontotemporal Lobar Degeneration (FTLD) is a group of brain disorders that have apraxia of speech as its earliest and main symptom. The prediction of the underlying physiopathology of these diseases, which prevalence is increasing each year, is a current challenge for health researchers.


Our main goal is to provide tools that can help physicians in the assessment. To do that we intend to build upon the experiments available for English, to provide an accurate characterization of phonetic markers of FTLD speech in Spanish/Catalan focusing on which are the most reliable phonetic parameters for the detection of the disease and for the classification of the underlying FTLD type. We also will compute the minimum speech sample required in order to get a reliable value of the parameters studied. Our ultimate goal is proposing a fast speech test (a sentence or short text) that can be automatically analyzed by means of a tool that will be designed for the purpose.


We expect that, as happens for English, speech features can be a marker for at least some of the underlying FTLD types. Out of the acoustical measures, we expect that temporal measures (such and number and duration of pauses, speech rate and rhythmic metrics) are more reliable than spectral features and therefore more suitable for the automatic analysis. We also expect that classification results are better in spontaneous and long tasks; however, we hope to have reliable results with a relatively short speech sample. Finally, we expect women to have acuter cases of AOS being easier to distinguish groups in their data.


Participants: The participants of this study will be patients suffering from FTLD recruited by Hospital Sant Pau and a cohort of healthy controls. All the participants patients and controls undergo a series of medical tests (neurologic, cognitive and functional) in which a speech protocol has been included. The speech protocol includes the production of a sustained “a”, repetition of syllables, pseudowords, sentences, a reading of a phonetically balanced text and the description of a cookie theft sheet and a picnic scene sheet and a spontaneous speech simple, among others. The protocol is available for Spanish and will be adapted to Catalan.

Analysis: We will annotate de data manually in Praat and extract promising phonetic features (speech rate, number of pauses and duration, rhythmic metrics, pitch slopes and voice quality measures). The same parameters will be computed automatically without manual annotation. We will compare and analyze the results using a mixed-effects model or a Bayesian approach. We will benefit from the tools created in the automatic analysis to build a web-app that, given a preestablished sample of speech, computes the phonetic markers of FTLD found in this project.

Expected results

The most straightforward result is an annotated corpus of FTLD patients. But the main result of this project will be advancing in the knowledge of speech markers FTLD (both the acoustical measures and the minimum speech sample necessary to have significant results). Finally, this knowledge will serve to implement an app that can compute them automatically. In sum, we hope we can improve in-vivo prediction of underlying pathology in FTLD by adding acoustical analyses to the regular tests conducted in medical practice helping to give patients an early treatment that helps them to keep a good quality life longer.



Lorraine Baqué, PhD

Universitat Autònoma de Barcelona


Victoria Marrero Aguiar, PhD



Work team

Joan Borràs-Comes, PhD

Universitat de Barcelona

Sonia-Karin Marqués Kiderle, SLP

Hospital de Sant Pau


Sara Rubio Guerra, MD

Hospital de Sant Pau


Elisabet Vila, PhD

Universitat de Barcelona



Miguel Ángel Santos-Santos, MD, PhD

Hospital de Sant Pau


María Machuca, PhD

Universitat Autònoma de Barcelona


Findings and news

Language diversity also matters in clinical speech

PVI, a metric that works in English with 10 nfPPA patients for diff. nfPPA from semPPA does not reach as good results in Spanish, even with 21 patients. Read the paper to discover what makes Spanish rythm different from English. Publication

The Speech-FTLD is now adapted to Catalan

The adaptation of the protocol to Catalan has been finished and it is now being validated.

Segment duration is increased for all PPA groups compared to controls

Semantic variant of PPA is often used as control group since it is usually claimed that those patients do not have motor speech problems. However, our data shows that there is an increase on duration of 9.07ms (±8.37) for PPA-sem compared to controls. Publication

Related publications

Elvira-García, Wendy; Victoria Marrero-Aguiar, Joan Borràs-Comes, Miguel Ángel Santos-Santos. 2023. Speech metrics that differentiate primary progressive aphasia variants depend on language typology. International Conference of Phonetic Sciences (ICPhS). Praga, 7/08/2023 [Poster][Presenter: Wendy Elvira García] Download proceedings

Baqué, L. & Machuca, M. J. (2023). Hesitations in Primary Progressive Aphasia. Languages, 8(1), e45. Download

Baqué, L.; Machuca, M.J. Assessing dysfluency in Primary Progressive Aphasia: speech-related vs pause-related parameters. ICPLA'2023. Salzburg (Austria). 04/07/2023. [Oral communication]

Show previous publications

Elvira-García, Wendy; Victoria Marrero-Aguiar; Joan Borras-Comes; Sara Rubio-Guerra; Sonia Karin Marques-Kiderle; Núria Montagut; Carlota Faixa; Laura Auge; Camille Wagner Rodriguez; Stephanie M Grasso; Alberto Lleo; Ignacio Illán-Gala and Miguel A. Santos-Santos. 2023. Acoustic speech metrics that differentiate primary progressive aphasia variants depend on language typology. Alzheimer's Association International Conference. Amsterdam, July [Poster][ [Presenter: M. Ángel Santos]

Baqué, Lorraine, Machuca, Mª Jesús, & Santos-Santos, Miguel Ángel. (2022). Estudio preliminar de variables temporales del habla continua en pacientes con síndromes neurodegenerativos del espectro degeneración lobar frontotemporal / Preliminary study of the temporal variables of continuous speech in patients with neurodegenerative syndromes of the frontotemporal lobar degeneration spectrum. Revista de Neurología, 74(02), 37. . {ISSN: 1576-6578}

Elvira-García, Wendy; Marrero, Victoria; Bargiela Zotes, Víctor; Santos-Santos, Miguel Ángel 2022. La tasa de habla y la duración segmental en el diagnóstico de síndromes neurodegenerativos: Palaba vs. pseudopalabra. L Simposio - IV Congreso de la Sociedad Española de Lingüística. 7th April. Madrid, Spain.

Elvira-García, Wendy; Marrero Aguiar, Victoria. 2021. Syllable repetition speech rate and rhythm as a tool to distinguish types of neurodegenerative aphasias. Phonetics and Phonology in Europe (PaPE). Online.

Farrús, M.; Elvira-García, Wendy; Garrido, J. M. 2021. On the Need of Standard Assessment Metrics for Automatic Speech Rate Computation Tools. Phonetics and Phonology in Europe (PaPE); Workshop "From speech technology to big data phonetics and phonology - a win-win paradigm" Online. Download

Baqué, Lorraine (2020). How do persons with apraxia of speech deal with morphological stress in Spanish. A preliminary study, Clinical Linguistics & Phonetics, 34(1-2): 131-168. {Print ISSN: 0269-9206. Online ISSN: 1464-5076. DOI: 10.1080/02699206.2019.1622155}