Today, The Next Web talked about a patent received by Spotify, which describes a system for determining a user's mood based on audio data from a microphone. The patent was issued on January 12, but journalists were only able to find it today.
The published document describes the operation of an algorithm that takes as a basis the audio data of the user's speech and background noise. Based on this data, the application will determine the emotional state, gender, age, and race of the user by accent. To determine, the algorithm will analyze the pitch, timbre, and rhythm of the voice. Apart from this, the application will also be able to recognize stress or bad mood. The patent describes several application scenarios: one of them says that the information received will affect the output of subsequent tracks when the function of selecting similar music is enabled if the composition is over. In another, we are talking about the algorithm for forming playlists and also creating a dedicated section on the main page.
It is not known how the mood-based recommendation algorithm will select the music. Will Spotify play sad songs if the user is in a bad mood? Or, on the contrary, will it include encouraging motives? And if you are in a good mood - will Spotify support it with recommended music or make you think about something high at some point? In general, there are still many undisclosed details.
It is worth noting that this patent may not be implemented in the application itself, but it gives an opportunity to look into the near future and understand how the next generation recommendation algorithms will work. You can familiarize yourself with the patent on the official website of the Music Business Worldwide magazine.
0 Comments