Smart Voice A voice analysis app for sustainable health monitoring

Project overview

The aim of this project is to develop a machine learning-based smart mobile app and cloud-based infrastructure for the analysis of voice recordings, to non-invasively assess the presence of diseases, such as recurrent respiratory papillomatosis (RRP), in rural areas where the availability of trained medics and health professionals, and of technical capabilities is limited.

Recurrent respiratory papillomatosis (RRP) is a condition where wart-like growths occur in the larynx. While being a benign condition, RRP can spread through the airways and can on occasions become malignant.

More worryingly, when untreated, RRP can spread and actually obstruct the airways, with fatal consequences. The vocal folds are usually the first and most prominent site of any RRP growth, causing children to have a hoarse voice quality and/or stridor (a high-pitched, wheezing sound).

Children with voice problems are routinely assessed in a generalist ear, nose and throat clinic. Differential diagnosis of voice problems requires detailed case history, evaluation of voice quality along with visualisation of the larynx. This is done wherever possible with a small endoscopy camera inserted through the nostril. Not all children can tolerate this in the outpatient clinic so further investigation is necessary under general anaesthetic.

In developing countries, RRP can have a more complex presentation. It is known that there are considerable difficulties in primary diagnosis, attributed to lack of endoscopic equipment in rural areas.

Prevalence of RRP in developing countries is reportedly much higher than for the UK. In Africa and Central/South America, over 70% of children who present with RRP go on to require a tracheotomy. This involves the insertion of a tube directly into the trachea (windpipe) so that breathing can be continued safely through bypassing the larynx.

This is usually the case due to late diagnosis of RRP. Understandably, the tracheotomy procedure itself presents with further potential medical complications, with implications for supporting these children.

What’s involved for students?

The types of project activities that you could get involved with include (but are not restricted to):

• carrying out a literature review on similar projects or initiatives, and on similar mobile apps currently available
• carrying out a literature review on signal processing and machine learning techniques for voice analysis
• validating and optimising the relevant voice analysis techniques developed by researchers at Strathclyde
• developing a mobile app, to record voice data samples
• incorporating such voice analysis techniques into the mobile app developed
• exploring the viability of cloud-based solutions for voice analysis
• developing a cloud-based system for uploading voice data samples
• developing a database system for storing voice data samples and relative analysis outputs
• developing a mobile app to access the cloud-based database of voice data samples and analysis reports
• developing a business model for the deployment of the mobile apps developed
• preparing marketing material for the system developed

Because of these varied tasks, we encourage all interested parties to get in touch, even if your department is not listed above.

Please note: availability can vary between degrees. Please contact your advisor of studies and the project lead for more information.

Have more questions, or want to get involved?

Contact either:

• gaetano.di-caterina@strath.ac.uk
• wendy.cohen@strath.ac.uk 

Apply for this project