BLOG: “AVATAR” – digital technologies for children’s speech sound disorders

Thanks to Rebecca von der Ley for sending us this blogpost to describe her experiences of working on a research project.

Hello, my name is Rebecca. I have a BSc in SLT and I am currently studying Evidence-based Health Care (MSc) at the Hochschule für Gesundheit (hsg) in Bochum, Germany.

I work part-time as a research assistant in the project “AVATAR” at the hsg and I would like to tell you a little bit about this project:

The third-party funded project “AVATAR” runs for three years.

It involves SLTs, software developers, researchers and illustrators, on the topics of computer vision machine learning, cognitive signal processing.

The aim of this project is to develop a technical assistance system to support speech therapy exercises for children aged 4-6 years with isolated articulation disorders.

“AVATAR” is a pioneer in the field of providing speech assistance via up-to-date audiovisual speech recognition.

The application can be used to deepen speech therapy treatment in the context of home-based training for children. It intends to increase therapy efficiency and to shorten therapy duration by means of individualised exercises and documentation for SLTs.

In Germany, more than 10% of the total sales of medical products include speech therapy services.

Therapy of children and adolescents amounts to approximately 65% of the speech therapy services of which about 50% due to articulation problems.

The therapy of articulation disorders often takes a long time. It requires children to practice intensively and sustainably at home.

It has been proven that exercises at home have a positive effect on the success of therapy, but it is often difficult for SLTs to offer motivating exercises that are appropriate for the children.

Against this background, media-based diagnostic and therapeutic interventions will play an important role in the future to ensure the quality of therapy and to reduce healthcare costs.

Therefore, the goal of the “AVATAR” project is to develop a computer program or an app which will recognise the child`s articulation due to an individual audiovisual feedback. Additionally, it supports correct forming.

Through the program, the child’s perception is going to be trained. The exercises are developed as games and are adjusted to the children`s interests.

“AVATAR” provides following advantages for SLTs and users:
* The program provides information on how constant and successful the practice has been.
* It gives timely feedback on articulation and individualised assistance to the user.
* It can be practised independently at home with high motivation.

Automatic speech recognition has made rapid progress in the last years. Current speech recognition systems are adjusted to the recognition of regular adult speakers’ language.

Due to an exceedingly higher availability of training material, they are more advanced in the English language than in the German language.

To train speech recognition for German children, the SLTs are recording and annotating the language as one of many of their competencies in the project. In this process, the children have to speak 70 words on the topic of “farm”.

Microphones record the spoken words and cameras film their mouth and tongue movements. The collected data material is used to train a speech recogniser by researchers.

Here is what I feel I have learned from this experience:

– I got an insight into project management
– I learned to work in a structured and independently manner
– I learned to transact evaluations and to analyse them statistically
– I gained new knowledge about artificial intelligence and neural networks
– I practised dealing with participants
– I expanded my knowledge in articulation disorders
– I did systematic literature research
– I improved my work with a program for sound processing and annotating