Result description
We developed a method that is able to detect emotions in users of a smartphone application. To do this, we used audio data from the speech of the users. We do not analyze the content of what the users are saying, but the intonation and the way they express themselves, which contributes the users’ privacy. The recordings are removed from the server as soon as features for speech emotion recognition have been extracted from them. We use machine learning, with an emphasis on good performance in new situations by means of transfer learning. The recognised emotions can be used to improve interaction with the smartphone application, understand the mood of the users, and other purposes.
Addressing target audiences and expressing needs
- Grants and Subsidies
- Business partners – SMEs, Entrepreneurs, Large Corporations
- Technology Transfer Expertise
We are interested in further development of our result in research/development projects. We seek to increase accuracy, extend scope and similar.
Alternatively, we are interested in making this result more mature and better adapted to commercially relevant applications. This could be done to satisfy the needs of a specific customer (who would fund the adaptation), or for general commercialisation in partnership with a company.
- Public or private funding institutions
- Other Actors who can help us fulfil our market potential
- Research and Technology Organisations
R&D, Technology and Innovation aspects
Our software has been integrated in the WellCo virtual coach prototype. It works well in this context. Next steps can be in several directions: (1) make minor refinements to incorporate in a commercial version of the WellCo system. (2) Identify commercial applications in need of such software, and adapt it to them. (3) Identify datasets to which the developed method can be applied, and refine it in research context. Several of these steps can be done in parallel.
Result submitted to Horizon Results Platform by INSTITUT JOZEF STEFAN
