Spotify patent hints at speech recognition for music recommendations

Digital music service Spotify has been granted a patent titled “Identification of Taste Attributes From an Audio Signal” for the invention of a system to use speech recognition and analysis to create individually recommended music choices, reports Forbes.

The patent was filed in February 2018, and granted last month.

Using personal voice data such as age, gender, emotional state and accent combined with the users’ physical environment (like whether they are on a train, at a party, or at school), the content matching algorithm is able to generate a list of recommendations taking into account previous listening history and friends or connections’ tastes.

“Speech content might be processed to eliminate words or phrases below a particular confidence level,” according to the patent.

The speech data is not used as a biometric to uniquely identify the individual during the process, and Spotify also gives users the option to enter their gender and date of birth at sign-up.

It is not clear when and how this new system would be implemented, and a spokesperson for the company hinted it may not come into use at all; “Spotify has filed patent applications for hundreds of inventions, and we regularly file new applications,” the representative wrote. “Some of these patents become part of future products, while others don’t. Our ambition is to create the best audio experience out there, but we don’t have any news to share at this time.”

Spotify is used by 320 million people across 92 markets, and it is not yet clear, should the invention be used, whether listeners will have a choice to opt out of the service.

Article Topics

biometric data  |  biometrics  |  personalization  |  research and development  |  speech recognition