Where speaking with Siri was once a forced, awkward endeavor, we’ve now become accustomed to talking to our devices. Whether we’re asking for the World Series schedule or would like Alexa to read us a recipe, human-to-device communication is no longer unusual. Still, as new such devices appear in the marketplace, there are some potential privacy concerns as we explore their capabilities.
Can You Hear Me Now?
Recently, more devices have appeared that are always (or almost always) on—always listening to what’s happening around them. The GoPro Hero 6 is one example. A largely voice-controlled camera, a user can instruct the GoPro to begin recording, even when the camera is off. Though you can change the setting, it is also possible for your camera to remain in a low-power listening mode that allows only the microphone to remain on. But it’s on all of the time, lasting up to eight hours and draining between five and 10 percent of the battery.
And the GoPro certainly isn’t alone.
New and Different
CEVA and Cyberon recently announced a partnership which offers a low power voice activation solution that is interestingly always listening. Designed for smartphones, consumer, and IoT devices, the CSpotter is Cyberon’s optimized implementation of their voice recognition engine for the CEVA-TeakLite family of DSPs. The CSpotter on CEVA-TeakLite-4 silicon will be demonstrated on October 27 at the CEVA Technology Symposium in Taiwan.
The CSpotter listens to ambient speech to detect and respond to predetermined words and trigger commands. It supports 33 languages based on phoneme acoustic models and has optimized both speaker independent and dependent voice activation. Currently, it already exists on CEVA-powered merchant silicon.
“The CEVA-TeakLite family of DSPs already powers billions of smartphones, consumer electronics, and other smart IoT devices, making CEVA an ideal partner for our CSpotter technology. CSpotter on CEVA’s DSPs is a highly compelling offering, with unmatched accuracy and performance in the noisiest environments and at a minimal power consumption,” said Tai-Hsuan Ho, Cyberon founder and CEO.
Moshe Sheier, who serves as the director of strategic marketing at CEVA, believes that Cyberon’s low footprint voice trigger and word spotter, as well as the offered support across a variety of languages makes the CSpotter something special. “Coupled with the world’s lowest power audio/voice DSP family, the joint implementation offers a perfect combination of performance and power efficiency for adding an always-listening voice interface to a wide range of smart devices, including smartphones, toys, wearables, automotive, and IoT,” he said.
An Industry Shift?
Some have suggested a shift in the industry toward a voice-controlled relationship with technology, which sounds like an easy and elegant solution for human-to-device communication. But what about security?
If enterprise systems begin relying on voice commands, a number of privacy concerns arise. First, who (and what voices) will be able to control said devices? Secondly, what happens to the recordings? And finally, who, if anyone, can tune in to listen or in some cases view the always-on devices?
Image courtesy of Pixabay.
Addressing Privacy Concerns
Some companies have begun to address these concerns. Google, for example, recently introduced a new camera geared toward parents that can constantly record. The Clips camera uses machine learning to begin to identify subjects you’d like photos of. By setting it on a counter, you can walk away and return to a host of moments that the camera has determined might be interesting to you, all recorded as snapshots for your viewing pleasure. The images are only accessible via your device and aren’t communicated to the cloud, which makes some users feel safer about a perceived lack of privacy.
Undoubtedly, many companies will have to develop strategies to combat consumer privacy fears as voice technology continues to develop. For now, users are perhaps encouraged to enjoy the changing landscape of technology.