Ambient Intelligence (AmI) is an interdisciplinary research area of ICT which has evolved since the 90s, taking great advantage from the advent of the Internet of Things (IoT). AmI creates, by using Artificial Intelligence (AI), an intelligent ecosystem in which computers, sensors, lighting, music, personal devices, and distributed services, work together to improve the user experience through the support of natural and intuitive user interfaces. Nowadays, AmI is used in various contexts, e.g., for building smart homes and smart cities, providing healthcare, and creating an adequate atmosphere in retail and public environments. In this paper, we propose a novel AmI system for gym environments, named Gym Intelligence, able to provide adequate music atmosphere, according to the users’ physical effort during the training. The music is taken from Spotify and is classified according to some music features, as provided by Spotify itself. The system is based on a multi-agent computational intelligence model built on two main components: (i) machine learning methods that forecast appropriate values for the Spotify music features, and (ii) a multi-objective dynamic genetic algorithm that selects a specific Spotify music track, according to such values. Gym Intelligence is built by sensing the ambient with a minimal, low-cost, and non-intrusive set of sensors, and it has been designed considering the outcome of a preliminary analysis in real gyms, involving real users. We have considered well-known regression methods and we have validated them using a collected data (i) about the users’ physical effort, through the sensors, and (ii) about the users’ music preferences, through an Android app that the users have used during the training. Among the regression methods considered, the one that provided the best results is the Random Forest, which predicted Spotify music features with a mean absolute error of 0.02 and a root mean squared error of 0.05. We have implemented Gym Intelligence and deployed it in five real gyms. We have evaluated it conducting several experiments. The experiments show how, with the help of Gym Intelligence, the users’ satisfaction about the provided background music, rose from 3.05 to 4.91 (on a scale from 1 to 5, where 5 is the maximum score).
Providing music service in Ambient Intelligence: experiments with gym users
Roberto De Prisco;Alfonso Guarino;Nicola Lettieri;Delfina Malandrino;Rocco Zaccagnino
2021-01-01
Abstract
Ambient Intelligence (AmI) is an interdisciplinary research area of ICT which has evolved since the 90s, taking great advantage from the advent of the Internet of Things (IoT). AmI creates, by using Artificial Intelligence (AI), an intelligent ecosystem in which computers, sensors, lighting, music, personal devices, and distributed services, work together to improve the user experience through the support of natural and intuitive user interfaces. Nowadays, AmI is used in various contexts, e.g., for building smart homes and smart cities, providing healthcare, and creating an adequate atmosphere in retail and public environments. In this paper, we propose a novel AmI system for gym environments, named Gym Intelligence, able to provide adequate music atmosphere, according to the users’ physical effort during the training. The music is taken from Spotify and is classified according to some music features, as provided by Spotify itself. The system is based on a multi-agent computational intelligence model built on two main components: (i) machine learning methods that forecast appropriate values for the Spotify music features, and (ii) a multi-objective dynamic genetic algorithm that selects a specific Spotify music track, according to such values. Gym Intelligence is built by sensing the ambient with a minimal, low-cost, and non-intrusive set of sensors, and it has been designed considering the outcome of a preliminary analysis in real gyms, involving real users. We have considered well-known regression methods and we have validated them using a collected data (i) about the users’ physical effort, through the sensors, and (ii) about the users’ music preferences, through an Android app that the users have used during the training. Among the regression methods considered, the one that provided the best results is the Random Forest, which predicted Spotify music features with a mean absolute error of 0.02 and a root mean squared error of 0.05. We have implemented Gym Intelligence and deployed it in five real gyms. We have evaluated it conducting several experiments. The experiments show how, with the help of Gym Intelligence, the users’ satisfaction about the provided background music, rose from 3.05 to 4.91 (on a scale from 1 to 5, where 5 is the maximum score).I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
