Our behavior is driven by emotions.

In many contexts, such as call centers or telephone calls, the interaction between two individuals is focused only on the voice and the emotions of the speaker can be uniquely inferred from his vocal expressions. Thus, defining a system that can understand people's emotions from speech and adapt accordingly is becoming a significant problem.

Here is the concept of Speech Emotion Recognition or SER for short. In fact, Speech Emotion Recognition systems try to recognize the emotional state of the speaker through the processing and classification of vocal signals. 

One of the goals of the ampel project is to develop a model that is able to recognize emotions, especially in the elderly. Through the "filo d'argento" service designed by Auser, i.e. a telephony service in support of fragilities and in contrast to loneliness, we record the voice of elderly people to create a dataset.

The lack of data makes it hard to implement reliable SER systems when elderly Italian people are taken into account. As a result, there is an urgency for a more realistic collection of data, that faces the need to consider a more heterogenous population of individuals having different ages or speaking different languages. The SER-AMPEL dataset was born with the idea of addressing this shortcoming. The dataset will be collected from a real-world scenario and will consist of audio recordings of natural conversations between volunteers and subjects. We hope that these new data could increase our knowledge on the topic of emotion recognition, thus giving significant help in the definition of robust and accurate systems able to distinguish affective states also in the case of people with different languages or ages.