The HUME
The HUME is an emotion recognition platform based on sensors, behavioral models and machine learning. The sensors measure physiological characteristics such as heart rate, skin conduction and activity. This information is converted into usable information in the HUME database by means of behavioral models. The HUME database is populated by a group of respondents we analyzed in our HUME lab.
The features
- Early warning in case of escalation
The HUME warns the caregiver in advance when stress builds up in a client. The caregiver can then intervene sooner and limit the consequences for himself or herself, but also for the client, such as self-mutilation. - Diagnostic function
Measurements with the HUME help to investigate and better understand the cause of escalations or special behavior. - Effect measurement
The HUME gives more insight into the effect of an intervention in approach, offer of activity, change in medication etcetera. - Warning of under alertedness
The HUME keeps an eye on a client at any time, it also notices under alerdness. In case of too long periods of under-stimulation, the caregiver will be warned so that it will have no effect on day and night behavior.
Validation
We have done measurements at more than 10 care institutions during the HUME pilot study. All these partners can be found below. During these pilot studies we have validated our models. Through machine learning and personalization these models can perform even better, giving HUME an accuracy of >80%.
Publications
- Real-Time Arousal Detection Based on Skin Conductance and Heart Rate Features
- Analysis of physiological signals for recognition of stress
- Emotion sensing to improve quality of life of people with a mental disability
Real-Time Arousal Detection Based on Skin Conductance and Heart Rate Features
Analysis of physiological signals for recognition of stress
Emotion sensing to improve quality of life of people with a mental disability