Adaptive intelligent speech processing system of medical personnel with the structuring of test results and support of therapeutic process

The project aims to develop and implement a solution by means of which physicians will be able to voice-recall available diagnostic test results and clinical parameters of patients, fill in disease charts during medical history taking, create radiological descriptions, and prescribe treatment as required. The system to be developed will automatically generate templates for their completion, including medical history, radiology image descriptions, which will be editable by voice, and will also enable dictation of referrals for additional tests, prescriptions, and sick leave. The cloud-based speech recognition system will be built on the basis of the developed corpus of Polish speech extended by the dictionary of medical terms and trade names of medicines, intended to be used primarily by doctors of various specializations, including radiologists, surgeons, doctors working in hospital emergency wards, specialists giving medical advice. The variant of the solution extended with two-way communication, based both on speech recognition and speech synthesis, will be used when a doctor cannot operate a text editor manually. It is assumed that the corpus of recordings will be based on recordings of articulated speech in typical acoustic conditions of doctors' offices and also in surgical masks, in operating theatres and also in conditions that make effective speech recognition difficult, i.e. in the presence of reverberation, interference and simultaneous speech. In particular, to use the system in operating theatres and clinical emergency departments, the constructed hardware solution for speech signal acquisition will be used. An additional application of the system will be to search medical databases and repositories. Hence the repositories will be supplemented with the names of medical conditions and some drug components as well as trade names of popular medical products. Special attention will be paid to training the system for recognition of associated sets of words, elimination of repetitive erroneous terms, dictation of dimensions of radiological lesions, voice correction of entries in referenced templates, ability to read entries using synthetic speech, starting from highlighted cursor positions and replacement of existing fragments with new ones dictated by voice.

Details

Project's acronym:

ADMEDVOICE

Financial Program Name:

INFOSTRATEG

Organization:

Narodowe Centrum Badań i Rozwoju (NCBR) (The National Centre for Research and Development)

Agreement:

INFOSTRATEG4/0003/2022 z dnia 2023-05-04

Realisation period:

2023-04-01 - 2026-03-31

Project manager:

prof. dr hab. inż. Andrzej Czyżewski

Team members:

Expert prof. dr hab. inż. Bożena Kostek
Machine learning specialist dr hab. inż. Piotr Szczuko
Acoustic systems designer dr hab. inż. Józef Kotus
Database designer dr hab. inż. Grzegorz Szwoch
Test automation specialist dr inż. Piotr Odya
Voice biometrics specialist dr inż. Maciej Szczodrak
Cloud technology specialist dr inż. Arkadiusz Harasimiuk
Specialist dr inż. Karolina Marciniuk
Interface development specialist mgr inż. Paweł Spaleniak
Speech recognition specialist mgr inż. Szymon Zaporowski
Machine learning specialist dr inż. Sebastian Cygert
Specialist mgr inż. Dawid Weber
Specialist mgr inż. Wanda Ludwikowska
Specialist mgr inż. Franciszek Górski
IT technician Daniel Wiśniewski
Machine learning specialist inż. Marta Zielonka

Realised in:

Department of Multimedia Systems

External institutions participating in project:

Gdański Uniwersytet Medyczny (Poland)
E-TRUST Sp. z o.o. (Poland)

Project's value:

8 299 875.00 PLN

Request type:

National Research Programmes

Domestic:

Domestic project

Verified by:

Gdańsk University of Technology

Papers associated with that project

results on page:
year:
- selected sort by year newest first
- sort by year oldest first
title:
- selected sort by title A-Z
- sort by title Z-A
citation:
- sort by citation descending
- sort by citation ascending

Filters

total: 3

Year 2024

Enhancing voice biometric security: Evaluating neural network and human capabilities in detecting cloned voices
Publication
- A. Czyżewski
- Journal of the Acoustical Society of America - Year 2024
This study assesses speaker verification efficacy in detecting cloned voices, particularly in safety-critical applications such as healthcare documentation and banking biometrics. It compares deeply trained neural networks like the DeepSpeaker with human listeners in recognizing these cloned voices, underlining the severe implications of voice cloning in these sectors. Cloned voices in healthcare could endanger patient safety by...

Full text to download in external service
The Impact of Foreign Accents on the Performance of Whisper Family Models Using Medical Speech in Polish
Publication
- S. Zaporowski
- Year 2024
The article presents preliminary experiments investigating the impact of accent on the performance of the Whisper automatic speech recognition (ASR) system, specifically for the Polish language and medical data. The literature review revealed a scarcity of studies on the influence of accents on speech recognition systems in Polish, especially concerning medical terminology. The experiments involved voice cloning of selected individuals...

Full text available to download

Year 2023

A survey of automatic speech recognition deep models performance for Polish medical terms
Publication
- Year 2023
Among the numerous applications of speech-to-text technology is the support of documentation created by medical personnel. There are many available speech recognition systems for doctors. Their effectiveness in languages such as Polish should be verified. In connection with our project in this field, we decided to check how well the popular speech recognition systems work, employing models trained for the general Polish language....

Full text to download in external service

seen 10350 times