 |
Joana Correia, Francisco Teixeira, Alberto Abad, Catarina BotelhoINESC-ID / IST, University Lisbon, Portugal |
Isabel Trancoso is a full professor at IST (University Lisbon), and the President of the Scientific Council of INESC-ID. She received the Licenciado, Mestre, Doutor and Agregado degrees in Electrical and Computer Engineering from IST in 1979, 1984, 1987 and 2002, respectively. She was the President of the Electrical and Computer Engineering Department of IST. She was elected Editor in Chief of the IEEE Transactions on Speech and Audio Processing, Member-at-Large of the IEEE Signal Processing Society Board of Governors, and President of ISCA (International Speech Communication Association). She chaired the INTERSPEECH 2005 conference. She chaired the IEEE James Flanagan Award Committee, and the ISCA Distinguished Lecturer Selection Committee. She was a member of the IEEE Fellows Committee, the IEEE Publication Services and Products Board Strategic Planning Committee, and Vice-President of the ELRA Board. She currently integrates the ISCA Advisory Council, and the ISCA Fellow Selection Committee, and she chairs the Fellow Evaluation Committee of the Signal Processing Society of IEEE. She received the 2009 IEEE Signal Processing Society Meritorious Service Award. She was elevated to IEEE Fellow in 2011, and to ISCA Fellow in 2014.
Speech as a (private?) biomarker for speech affecting diseases
Speech has the potential to provide a rich biomarker for health, allowing a non-invasive route to early diagnosis and monitoring of a range of conditions related to human physiology and cognition.
The range of diseases that affect speech is much larger than the
so-called speech & language disorders (e.g. sigmatism, stuttering). In fact,
speech is affected by several diseases that concern respiratory organs, such
as the common Cold, or Obstructive Sleep Apnea, and may also reflect mood
disorders such as Depression, and Bipolar Disease. Most importantly, speech
may be an early biomarker of neurodegenerative diseases such as Parkinson’s,
Alzheimer’s, and Huntington’s disease.
With the rise of machine learning applications over the last decade, there
has been a growing interest in developing speech-based diagnostic tools.
Such tools are typically trained with limited speech data collected in
controlled conditions from patients diagnosed with a given disease and
healthy subjects, their output being a global indicator of whether a subject
is suffering from that disease.
Despite this recent progress, the potential of speech as a biomarker for health is very far from being explored. This potential may be addressed from different perspectives: on one hand, one may take a holistic view of speech as a biomarker for many diseases, instead of a specific one; on the other hand, one may also attempt to discriminate which aspects of speech are affected – e.g. it may quantify the degree of nasalization (relevant for the progress of Amyotrophic Lateral Sclerosis), the voice tremor (relevant for Parkinson’s Disease), the duration of breathing cycles in speech, etc.. Such discrimination may be relevant for speech therapists.
Another perspective is the robustness of this type of analysis for in-the-wild speech data, instead of lab data recorded in very controlled conditions. This aspect may be very relevant for making the analysis of speech just as common as blood tests are nowadays for medical diagnosis.
Last but not least, one should also address the privacy issues raised by
the potential use of speech as a biomarker for health conditions, namely
when this type of analysis is performed in remote servers. This may be done
by exploring the use of cryptographic techniques for privacy-preserving
machine learning.