How DSP Concepts helped design Spotify’s Car Thing, leveraging DSP’s experience with GoPro to filter wind noise, isolating voice commands from music, more (Janko Roettgers/Protocol)

how dsp concepts dsp goproroettgersprotocol Good morning, and welcome to Protocol Next Up. This week: how a California…

how dsp concepts dsp goproroettgersprotocol

Good morning, and welcome to Protocol Next Up. This week: how a California startup helped Spotify build Car Thing, and what’s next for Google’s Android TV and smart home platforms.

Making hardware is hard, especially for companies with no prior expertise in consumer electronics. So when Spotify set out to develop Car Thing, its new automotive display device, it got help from outside experts.

To make voice recognition for Car Thing work, Spotify relied on DSP Concepts, a Santa Clara-based startup that has built a software platform for audio hardware development. DSP Concepts CTO Paul Beckmann recently filled me in on how his team helped turn Car Thing from an idea into a real thing.

  • DSP Concepts worked on Car Thing for roughly 12 to 15 months, Beckmann said. “We like to get involved early,” he told me.
  • The company was tapped to help develop the device’s voice recognition, which includes what Spotify calls “adaptive interference cancellation” — algorithms that make sure Car Thing listens to voice commands, and not the lyrics of the song that’s playing, backseat chatter, engine sounds or freeway noise.
  • DSP Concepts has been working on audio for carmakers like Tesla, Mercedes and BMW, giving the company a pretty good idea of how to deal with these environmental sounds.
  • Beckmann and his team were also able to provide input on the placement of the four microphones used by Car Thing for hands-free voice control.

However, Car Thing is very different from your typical in-car entertainment system. The device functions as an interface for the Spotify app on your phone.

  • Car Thing makes it easier to pick songs, playlists and podcasts while keeping eyes on the road, but the phone is still providing connectivity, and sending the audio via Bluetooth or line-in to the car stereo.
  • That setup makes isolating voice commands, and filtering out noise, a lot harder. “It doesn’t have access to the music signal,” Beckmann said.

A carmaker also has complete control over the positioning of microphones and the acoustic qualities of each cabin component. Car Thing, on the other hand, is very much a DIY solution; consumers decide where they want to place it, and the device needs to work with whatever car they own, no matter how old or banged-up it is.

  • In essence, Car Thing needs to work more like a smart speaker, which consumers may place in many different environments. That’s why DSP Concepts did extensive recording and testing in real cars to optimize the signal processing.
  • On the flip side, smart speakers need to optimize for voice commands uttered from across the room. In a car, the person requesting the next song is never more than a few feet away. “Even American cars aren’t that big,” Beckmann said.

The biggest challenge: car dashboards, which are notoriously overcrowded. Among other options, Spotify ships Car Thing with a vent mount, and many consumers find that the vent grille really is the only option to add another screen to their car — which makes voice isolation all the more challenging. “The vent noise was surprising,” Beckmann told me.

However, DSP Concepts could rely on some past expertise to deal with blasting ACs. The company has been helping GoPro to add wind noise suppression to the company’s action cameras. Turns out that making audio work on a surfboard helps a lot when optimizing it on a dashboard.

Related Posts