By Miles Kingston
Recent advancements in artificial intelligence, machine learning and natural language processing are beginning to bring the full potential of smart homes into focus. As these technologies continue to learn and improve, many of the tasks of running the home will eventually be automated to provide peace of mind and enrich daily life. In fact, a recent Intel survey revealed 68 percent of Americans agreed that living in a home with smart devices would make their lives easier.1
Consumers are beginning to show an increasing readiness to welcome speech in the smart home. Look at smart speakers and personal assistants: This year, 35.6 million Americans will use a voice-activated smart home device at least once per month, eMarketer estimates – a 128.9 percent increase over 2016.
Intel, in collaboration with the Amazon Alexa Voice Service* (AVS), is making it easier for third-party developers to accelerate the design of consumer products featuring AVS. Today, Intel announces the Intel® Speech Enabling Developer Kit, which provides a complete audio front-end solution for far-field voice control.
Speech recognition has become a clear competitive advantage to product developers. Yet, giving machines the ability to listen, speak and converse in natural language is no easy feat.
Natural language means machines need to clearly recognize and respond to user commands from a reasonable conversation distance. People speak and hear in 360 degrees, not just in a direct line of sight. Devices need array microphones and complex noise mitigation technology. A quality voice interaction means devices identify the speaker’s location, mitigate and suppress ambient noise, and understand spoken commands on the mics, even while playing music (talking and listening at the same time), as well as waking up when it hears the wake word (e.g. “Alexa”).
There’s a lot of engineering involved in getting speech recognition at high degrees of speed and accuracy to deliver the best customer experiences. The Intel Speech Enabling Developer Kit is based on a new architecture that delivers high-quality far-field voice even in the most acoustically challenging environments. This marks the latest in a string of smart home innovations, including the Intel-powered Amazon Echo Show*.
The Intel® Speech Enabling Developer Kit will be available for pre-order now. Among the developer kit’s technology components:
- High-performance algorithms for acoustic echo cancellation, noise reduction, beamforming and custom wake word engine tuned to “Alexa”
- Intel’s dual DSP with inference engine
- Intel 8-mic circular array
With increasing speed, the innovations needed to power the future are within reach. In the smart home, we can all look forward to a wave of innovation from the developer community as we transition from being simply connected to being truly smart.
Miles Kingston is the general manager of the Smart Home Group at Intel Corporation.
1The Intel online survey was fielded among 2,552 Americans, 18+ in March 2017 to understand Americans’ perspectives about computer usage, personal technology, smart home technology, virtual reality and gaming. The survey was fielded via Ipsos.
Teaching Devices to Talk: Algorithms for Far-Field Voice
Beamforming identifies the location of the speaker and then channels the mic input corresponding to that location. Beamforming also helps in noise mitigation and ambient noise suppression.
Auto-Echo Cancellation (AEC) suppresses the audio output coming out of speakers from interfering with the microphone. It allows for “barge-in” commands when music is playing. The AEC algorithm requires the speaker’s signal as a reference into DSP for audio suppression.
Keyword Spotting (KWS) detects the utterance of wake words – “Alexa” – and notifies the host system. Firmware implementation in the DSP enables low-power operation of wake on voice for personal assistants.