Relive the ROSE event!

Relive the ROSE event!

Watch the video
Best Practice

Affective Computing: recognising emotions with AI

Discover how Affective Computing can create new digital interactions, making them more natural and engaging through the recognition of emotions.


Affective Computing, also known as Emotion AI, concerns the development of systems capable of recognising, interpreting, processing and simulating human feelings and emotions.

Interpret emotions and respond consistently

Affective Computing systems use Machine Learning models for the classification of emotions through the analysis of video, voice and text. This technology allows us to recognise feelings during a conversation (for example anger, disgust, fear, joy, sadness and surprise) and, applied to humanoid robots and Digital Humans, allows them to understand and respond to the emotions of the interlocutor in a more engaging and natural way. Thanks to Affective Computing, robots and Digital Humans can thus adapt their responses and reactions based on the emotional attitude of the interlocutor, thus improving interaction and communication.


From digital humans to robots

The applications that Reply is experimenting with in the field of Affective Computing are aimed at enabling more natural and engaging human-machine interactions. Here are two examples of this technology in action.


Ameca, the humanoid robot

Ameca is an extremely advanced humanoid robot. Thanks to Affective Computing, it offers an extremely realistic experience, reacting to changes in expression, changes in tone of voice and changes in language during interaction.


Rose, the digital human

Rose is Reply's digital assistant, an expert in technological issues. Thanks to Affective Computing, it is able to adapt responses based on facial expressions, changes in tone of voice and changes in the speaker's language, thus improving interaction.

Toward Deep Affective Computing

The next step in Affective Computing is Deep Affective Computing. This system uses a cognitive model composed of seven agents, each specialised in a single emotion, who propose solutions to the stimuli received. Next, a deterministic mathematical agent mediates between the seven solutions to maintain the system's balance and provide adequate answers.

In addition, the system evolves through interaction with the user, saving information such as their history, objectives and experiences. In this way, a cognitive system is obtained that, over time, learns to know the user in depth. In fact, this system uses the user's past experiences, preferences and objectives to provide personalised and relevant answers to their questions.


Try our Affective Computing solutions for personalised and engaging answers.


Machine Learning Reply is the Reply group company specialized in providing artificial intelligence services and solutions to guide its customers towards digitalization, helping them to become more competitive and guided by data thanks to Smart Analytics, Machine Learning and Artificial Intelligence. With experience in deep learning, computer vision, NLP and predictive modeling, the company helps its customers to empower their business, providing them with highly experienced dedicated development teams.

You may also like