Spotify is interested in using recordings of our voice and background noise to decide which music to recommend

It seems that Spotify not only want users to listen on their platform but their platform to listen to us. The company has been granted a technology patent that aims to use recordings of user voices and background noise to determine what type of music to select and recommend.

The name of the patent is "Identification of taste attributes from an audio signal", and describes a method and product for processing audio signals in which voices or background noise that is examined through speech recognition technology, to isolate dialog and environmental metadata from noise, and determine multimedia content preferences based on these .

Use our voice to determine our emotional state, gender, age or accent

Ethan Cull Gskx2swh74a Unsplash Blog

The patent describes the possible uses of the technology and talks about the extraction of "intonation, stress, rhythm and other similar elements from the speech units of the user's voice." And it goes so far as to imply that the Spotify application can even use that data to categorize the user's mood as "happy, upset, sad, or neutral".

That same technology could use voice recognition to identify metadata such as emotional state, gender, age, accent and even environment, i.e. if someone is alone or with other people, from the audio recording.

Discoverify is like Spotify's weekly Discovery but with new music every day

Spotify also explains in its patent how it currently uses a decision tree that shows users different artists and genres to help refine its user recommendation algorithm.

For now, the platform needs to ask the user things like their age or gender, and then to further refine the recommendations it has to offer options for the user to select artists they like, and according to Spotify the current method is not efficient enough.

This new technology would seek skip the step of the user providing the information:

What is needed is a totally different approach to collecting a user's taste attributes, particularly one that is based on technology so that the human activity described above (for example, requiring a user to provide information) is at least removed partially and performed more efficiently

At the moment the company has not commented on plans to implement the technologies described in this patent, and it is not uncommon for companies to register inventions that do not end up reaching the market. However, it is still interesting and perhaps on the alarming side at the level of user privacy. When do we accept as normal for an app to record our voice and environment just to recommend songs?

Via | Pitchfork