**5. Discussion**

The approach used to implement TherAImin could be used for simpler and more complex devices. As an example, the "Singing Elevator" can use the presence, temperature, humidity and noise sensors and additional information from the Internet as user stimuli. Using aggregation of estimators this process can be further improved [36]. Using a simple learning layer (local or not), it will decide from which category it should retrieve the generated music. The production layer would be a simple player with possible adaptations to handle noise in the cabin or other issues.

As a further example, the emotiWatch uses the wearable device sensors as a stimuli layer, preprocesses them in the stimuli adaptation layer, estimates the user's emotional state, and selects the music in the learning layer and outputs the music through a player in the production layer. The same approach can be used to start the design of any AIME.

It is clear that many other approaches could have been proposed, but FAIME is simple and gives clear insights into the musical device design process.
