Aug 11, 2015 | By Alec

Engineers are increasingly employing 3D printers to solve complex issues in the field of robotics, but we are still surprised by how simple these solutions sometimes are. This is again the case with a 3D printed solution for a problem almost as old as robots equipped with microphones themselves: how do you enable robots to distinguish between different voices? Something that is so simple for humans has proven to be exceptionally difficult for robots, having become known as the ‘cocktail-party problem’. However, with a very simple 3D printed solution, scientists from Duke University have now come up with a good, if somewhat unwieldy solution.

This ‘cocktail-party problem’ has been around in the artificial intelligence business for a while, and computers have simply struggled to continue to pick out your voice when nearby people are talking simultaneously. Some proposed solutions revolved around complex algorithms or equipping robots with a large number of microphones focused on various positions in the vicinity, but neither have proved very successful. But as they explain in an article in the Proceedings of the National Academy of Sciences, a team led by electrical engineers Steven Cummer and Yangbo Xie, 3D printing offers a fantastic solution.

What they have essentially come up with is a large thick, 3D printed plastic disk, featuring 36 openings on its side, and diverse honeycomb passages leading to a single microphone in the center of the disk. It is essentially a single-sensor listening system that combines acoustic metamaterials and compressive sensing techniques. ‘Different from previous research efforts that generally rely on signal and speech processing techniques to solve the “cocktail party” listening problem, our proposed method is a unique hardware-based approach by exploiting carefully designed acoustic metamaterials,’ they write. ‘We not only believe that the results of this work are significant for communities of various disciplines that have been pursuing the understanding and engineering of cocktail party listening over the past decades, but also that the system design approach of combining physical layer design and computational sensing will impact on traditional acoustic sensing and imaging modalities.’

So how does it work? Well each of the 46 passages to the microphone is unique and features subtly different ways of enabling sound to travel to the center, the scientists explain. Drawing a comparison to partially filled water bottles, air resonates with the sound of voices spoken over it, but the amount of water (the structure of the honeycomb) influences the exact frequency. All those voices can thus be distinguished from each other because the unique 3D printed shapes create variations that can be picked up by the single sensor. According to Yangbo Xie, we humans would not be able to distinguish the difference, but the algorithm used for the sensor can almost always tell which direction it comes from.

And how correct is it? ‘The device with a compact array of resonant metamaterials is demonstrated to distinguish three overlapping and independent sources with 96.67% correct audio recognition,’ they write in their article. This simple, but effective approach is already being hailed as an excellent solution. The only current issue would be its size – it is approximately the size of a very thick pizza – but there is obviously plenty of room for optimizing that design for further applications. The Duke University scientists have already said it could be applied to hearing aids and other acoustic imaging and sensing applications, so we might see a much smaller version of this technology in the near future. It just shows what you can achieve with a bit of 3D printed prototyping.

 

 

Posted in 3D Printing Applications

 

 

Maybe you also like:


   





Leave a comment:

Your Name:

 


Subscribe us to

3ders.org Feeds 3ders.org twitter 3ders.org facebook   

About 3Ders.org

3Ders.org provides the latest news about 3D printing technology and 3D printers. We are now seven years old and have around 1.5 million unique visitors per month.

News Archive