With this concept, we propose a new way of doing skiing exercises!

This prototype lets the user take voice-controlled skiing lessons without the use of an instructor. It works by using audio recognition AI embedded in the skiing helmet. The user has a dialogue with their helmet which suggests different challenging or relaxing lessons. The system will reward the user with points and level-ups for completing lessons.

Choose your lesson on top of the mountain or on the ski lift, and have fun!

Ski down and follow the instructions to complete the exercise

Receive feedback at the bottom of the slope

? Try it out here: https://creator.voiceflow.com/prototype/6218a35312551b7c4d54be8c

Instructions: Click the link above and click the blue button to start the conversation. To talk press the spacebar, make sure your laptop has a working microphone. Try to talk loud and clear and use simple words for the best results. Our voice coach has some difficulty with certain names, so try a simpler name if the program does not recognize your own name (we use ?Maarten? in most of our tests).

? Or watch a demonstration here:

Findings after testing:

Insights:

• The AI had difficulty with recognizing other accents or names.

• The AI only recognized words when speaking very clearly and deliberately.

• The conversation style and timing made people have to actively think and remember options.

• Errors in the conversation made the flow end

• The conversation was distracting while skiing, in the test there were already too many variables for the participants to answer the questions.

• Faulty AI can trigger emotions and frustrate people

Implications:

• The conversation style should be less distracting and mentally demanding.

• The timing should be slowed down.

With this concept, we propose a new way of developing a better skiing posture!

The idea of the prototype is to use posture recognition to recognize skiing posture. This could be used by beginner skiers to improve their posture as they are often not aware of their own posture while skiing. The concept would use external cameras to film skiers going down the slopes. Beginner posture is classified as passive and backward leaning while the experts actively engage their body and lean forward.

? Try it out here: https://teachablemachine.withgoogle.com/models/TxVgWv__5/

Instructions: Click the link. On the righthand side of the screen, you see the prototype. Give access to your webcam to start. You can zoom in a little bit if you would like to see it bigger, but unfortunately, the platform does not allow fullscreen. Now take a few steps back from your webcam in order to see your full body on screen. Now try to take a sideways skiing pose, use the pictures as a reference. Make sure you face left! Mess around with the AI to see if you can get it to an ?expert? level! If you are in doubt whether you are doing it correctly, check out the video underneath.

? Watch a demonstration here:

Findings after testing:

Insights:

• The AI recognized the different postures

• The AI has difficulty recognizing other people than the input footage

• The AI has difficulty recognizing postures with a busy background

• The AI was very sensitive and reactive

• The participants didn?t know how to improve their posture

Implications:

• The AI should include different types of postures.

• The AI should differentiate the different elements of the posture: back lean, knee bend, arm angle, etc.

• The AI should give feedback on which elements of the posture should be improved and how.

• The posture analysis could be communicated in AR through the skiing helmet.

With this concept, we propose a new way of getting up after falling!

The idea of the prototype is to use sensors to detect when the skier has fallen down. In this prototype we show that if the skier falls down a red light turns on, indicating that the skier needs help getting up and might be hurt. As future development, the red light could be transformed into more advanced signals that function as a personal guide to help you get up on your own, and/or stimulate you to keep going.

? Watch a demonstration here:

Findings after testing: