Sensory releases fast and efficient cross-language speech recognition solution

Adding speech recognition to a product is not easy: assuming the device has the necessary microphones and chips to detect speech, building software to correctly convert input in one language into a stimulus is a huge hurdle for most companies. Today, Sensory officially released a solution called VoiceHub that promises to add multilingual natural language understanding to any product, allowing businesses without NLU expertise to deploy globally viable smart hardware at scale.

Using the web-based VoiceHub portal, developers with “no programming experience” can build the wake words, simple commands, and large natural language vocabularies needed for a given product, including regional customizations across multiple languages ​​of English, French, Mandarin, Portuguese, and Spanish. Once the model is built on the web, VoiceHub makes it as easy to download it to a test device as scanning a QR code. The model then connects with Sensory’s TrulyNatural on-device speech recognition software, which has the necessary large vocabulary and parsing capabilities to respond to voice requests in multiple languages.

The launch of VoiceHub is significant for technology decision makers because it could radically improve time to market and product performance for products with voice interfaces, enabling any business to start using NLU to replace computers with touch and/or keyboard input, or to reduce the need for human voice input services. It also avoids sharing data with Amazon, a major concern for potential competing companies that might be considering Alexa-based solutions.

Sensory notes that its software runs on STMicroelectronics’ sub-$70 ARM-based ST32MP1 Discovery Board, as well as Android and iOS devices, enabling developers to prototype working products “in minutes, not days.” Of note: The finished solution runs directly on the device, so it doesn’t require an internet connection to work properly, which means enterprise developers can determine how much voice-related data to share.

In a video made with a larger ST32MP1 evaluation board, Sensory demonstrates a custom coffee shop ordering platform that lets customers buy coffee, tea, or hot cocoa drinks from a microphone-equipped coffee machine. After saying "Hey Barista," the customer orders a specific drink and then says "Confirm" to place the order. The inexpensive hardware includes dual microphones with noise cancellation, allowing Sensory's software to correctly recognize voice commands at a distance of 10 feet, even in the ambient noise typical of a coffee shop.

Since opening a limited beta program in October 2020, Sensory says developers have successfully tested VoiceHub with cars, wearables, smart speakers, and smart home products. It hopes the final version will accelerate brand voice experiences and domain-specific voice assistants in a wider range of applications. In addition, because VoiceHub is tied to TrulyNatural’s large multilingual vocabulary, Sensory hopes the solution will be “truly conversational” and able to understand “millions of unique phrases” rather than just a few or two words.