DSP Concepts Announces TalkTo Microphone Processing Technology for Voice-Controlled Products

March 13 2019, 01:10
DSP Concepts has launched TalkTo, a new brand and a major expansion of its voice recognition technologies. TalkTo is a family of software-based Audio Front Ends for voice-controlled products. It uses advanced microphone processing and machine learning techniques to deliver highly reliable voice control, even in high-noise environments and with multichannel audio products. According to the company, TakTo will enable voice interfaces even in the most difficult classes of products, like TVs and soundbars.

With TalkTo, developers can now offer a satisfying, voice-based user experience in categories where it was previously impossible, such as set-top boxes, multichannel soundbars, kitchen appliances, and TVs. In these applications, voice recognition is challenging because the product’s microphones are subjected to a wide variety of loud, interfering sounds.

Most voice-controlled products can cancel the sound from their own speaker, but can do little or nothing to cancel the unpredictable noise sources in busy homes such as nearby TVs, air conditioners, appliances, and music coming from other devices. TalkTo is far more advanced. It continually monitors the environment and learns the character of any interfering noise, then adapts its noise cancelling to provide the clearest possible voice signal for local and cloud-based voice recognition engines. TalkTo is so sophisticated it can even cancel noise from multichannel sources such as soundbars - a task that completely overwhelms most voice control systems.

“We created TalkTo because consumers are now accustomed to the speed and convenience of a voice-based interface, and are demanding that user experience in other products throughout their home - from controlling what’s on TV to dimming the lights or setting the microwave without having to first wash your messy hands,” says Chin Beckmann, CEO of DSP Concepts. “This demand places great stress on product development teams trying to figure out how to configure and manage microphone arrays in these applications, where noise and interference can be extreme and unpredictable. By incorporating TalkTo and one of our recommended microphone topologies, design teams can deliver the user experience their customers are demanding.”

According to DSP Concepts, TalkTo variants are able to accommodate arrays from two to eight mics, and includes Adaptive Interference Cancellation (AIC) for industry-leading wake word recognition in smart speakers. For set-top boxes, the solution uses a circular array, and easily recognizes the wake word even in the presence of loud audio from a nearby TV or soundbar. TalkTo for soundbar includes DSP Concepts’ proprietary N-channel Acoustic Echo Canceller (AEC) and AIC to deliver reliable wake-word recognition, even at cinematic volume levels. 

For appliances, IoT and generic connected devices, TalkTo achieves clear voice signals even with a variety of mic array configurations and in a wide range of product form factors. And since the solution runs on small, low-power SoCs and DSPs, and uses on-demand noise-reduction and progressive wake-up to minimize current draw, it supports hearable and wearable products, still producing clean voice signals even in high-noise environments. For one of the industry's most difficult applications, which is TVs, DSP Concepts also confirms that TalkTo is able to deliver reliable far-field voice recognition even with linear two or four-mic arrays.
related items