As more consumers choose to upgrade their devices and purchase new tech, the adoption of smart speakers continues to rise. The product that once may have experienced significant hesitation from consumers, is now being implemented in households all over the country. While this new product can make life a lot easier, it does come with some privacy concerns. But nonetheless, there are solutions!
Amazon’s Alexa was one of the first smart speakers to hit the market and it has gained a lot of traction since. As its capabilities and functionalities grow, the line between machine and man blurs.
As of only a few days ago, Alexa can now deliver the news like a real human being. Goodbye robotic voice and hello professional newscaster. This update will allow Alexa to speak in a more natural way, putting emphasis on certain words and phrases and adding in various tones.
This tech is possible thanks to a text-to-speech (TTS) system, which learns new speaking styles from a few hours of training. Amazon’s neural TTS model has two components: a generative neural network that converts phonemes into spectrograms, and a vocoder that converts those spectrograms into continuous audio signals. This allows the AI system to distinguish elements os speech independent of a speaking style and unique to that style.
Outsmarting the Smart Speaker
While smart speakers continue to dazzle consumers with their increasing capabilities and human-like functions, they also draw concerns to privacy. Taping private conversations or “listening” while they’re not supposed to are major concerns that have been raised.
That’s why Bjorn Karmann and Tore Knudsen created Project Alias, which is a 3D-printed attachment that goes directly on top of your smart speaker. The Raspberry Pi-powered microphone and dual-speaker, located on the inside, create white noise that prevents the speaker from activating. The noise is disabled when the speaker hears its wake phrase.