Electronics for the first prototype

In this blog post, we take a look at the structure of our first functional prototype and the considerations behind the hardware selection.

Computer

For the initial tests and for quick commissioning, we rely on a Raspberry Pi. It offers a good balance of performance, flexibility, and community support. However, as extensive AI functions are integrated, such as for image and speech processing, the Raspberry Pi will likely be replaced in the long term by an NVIDIA platform that is specifically optimized for AI applications.

Visual Detection: Finding the Right Camera

For capturing the environment, we have tested several camera systems. The most convincing was a fisheye camera, as it captures the entire room in one image. With the help of a YOLO-based AI algorithm, the image is then intelligently cropped, so that relevant content can be extracted and processed, for example, for eye contact or gesture recognition. This avoids the vision AI having to recognize too large areas.

Language and Sound: Microphone Array & Speakers

To be able to engage in natural dialogue with users, a speaker and a microphone array were installed. The microphone array allows for targeted recording even in louder environments and is essential for speech recognition and processing. To soften the technical look, these components are behind fabric, giving the device a friendlier, more organic appearance.

Touchscreen with Display Controller: Intuitive Interaction

At the heart of human-machine communication is a touchscreen, which is controlled by a separate display controller. It serves not only for visual output but also enables simple interaction with the avatar. With this, we aim to create a user-friendly interface without unnecessary complexity for the users.

Enclosure: Technology Meets Design

All the electronics are housed in a 3D-printed frame, which in turn sits in a CNC-milled wooden enclosure. This combination allows for quick adjustments internally and at the same time provides a warm, inviting look: an important aspect, especially in the sensitive home environment.

Challenge: Flexibility in Prototyping

One of the biggest challenges: Accommodating everything while remaining flexible. As a demonstrator, the AvatarMediKi is constantly evolving. Components are swapped, functions expanded: this demands a modular, accessible design. Our team continuously works on mastering this balance between technical complexity and design aspiration.