Subscribe to my weekly newsletter here
Amazon’s Super Bowl commercial featuring an under-the-weather Alexa has been circulating on the Internet for a few days now. It’s a funny spot featuring Cardi B, Anthony Hopkins, and other celebrities filling in for a coughing Alexa.
Earlier this week, I wrote about Amazon’s potential healthcare patents. It turns out they’ve got a patent for the question millions will be asking after this commercial airs during the Super Bowl too:
Why isn’t Alexa responding to her name from the TV?
In a patent filed shortly before the Amazon Echo was first released in 2014, Amazon details “devices, methods, and systems for detecting wake words and audio commands that should be disregarded,” for example, “in a television or radio advertisement.”
The patent primarily describes a matching technique to avoid unintentional waking of voice-activated devices. Amazon records and stores clips of the commercial in the (AWS) cloud, and when Alexa detects a wake word or an audible command, it compares the audio of the detected wake word or command to the commercial snippet. If the detected audio matches the commercial, the device recognizes that the detected audio is part of the ad and ignores it. Additionally, the device may record and buffer portions of detected audio that precede and/or follow the detected wake word. The preceding and/or following portion(s) may be used to assist in identifying if the detected audio matches a stored audio sample, for example from the advertisement.
Amazon also recently published a blog post about this, detailing how “teams build acoustic fingerprints on-the-fly within our AWS cloud. When multiple devices start waking up simultaneously from a broadcast event, similar audio is streaming to Alexa’s cloud services. An algorithm within Amazon’s cloud detects matching audio from distinct devices and prevents additional devices from responding. The dynamic fingerprinting isn’t perfect, but as many as 80 to 90 percent of devices won’t respond to these broadcasts thanks to the dynamic creation of the fingerprints.”
For example, when Jimmy Fallon opens his monologue with a bit about Alexa, Amazon has algorithms in place that may recognize that multiple devices are being inadvertently woken up, preventing additional devices from responding to the command.
The patent also details another method, by which an advertisement may be configured to include an audio signal inaudible to human that indicates to a detecting device that the audio of the wake word or other audible command in the program is an inadvertent wake word or command and that the device should disregard that particular command. Think of a dog whistle or those “mosquito devices” that play a sound at ultra-high frequencies to scare away the youths. Amazon’s blog post doesn’t mention this technique, but it’s possible Amazon’s ad employs both this method and the cloud-based comparison method.
Also, shouldn’t having a celebrity voice substitute as Alexa’s voice (as in the commercial) be some sort of premium feature?