Siri could recognize user's voice patterns for identification in future iPhone or iPad
Apple is continuing to come up with ways to secure its devices while still making it as easy as possible to use, with one concept involving unlocking an iPhone or iPad and performing a Siri request, but only if the voice it hears matches that of its owner.
Granted by the U.S. Patent and Trademark Office on Tuesday, the patent for "Device access using voice authentication" is relatively straightforward, namely detecting the speaker's voice for a vocal request and determining if it is the main registered user.
According to the patent, a device capable to receive speech input from a user could conceivably create a "voice print" for the owner, using multiple examples of their speech. This text-independent voice print would determine characteristics of the user's voice to create a model, either on the device or via an external service, with the result being the point of comparison for future checks.
The voice print could use "statistical models of the characteristics" of a user's pronunciation of phonemes, to create the signature. In theory, this could include characteristics such as voicing, silences, stop bursts, nasal or liquid interference, frication, and other elements.
By being text-independent, the voice print would theoretically be able to determine the speaker's identity just from normal speech, without requiring the use of a specific passphrase. This would be a better way of dealing with voice-based authentication in public situations, where users may not wish to loudly state a passphrase they would rather keep private.
At the time the device receives a vocal command, it effectively creates a model based on the uttered phrase, which it then compares against the already-established voice print. If there is a certain threshold met for similarity, the device could be unlocked then carry out the spoken command.
It is also suggested a voice print match may not need to unlock the phone entirely, but could still enable the spoken request to be performed for the user. For example, Siri on a device could provide a verbal and text-based response to a vocal query about a general topic if it detects the user's voice, but would refrain from providing a response involving user data unless the device is unlocked by other means beforehand.
There is even the possibility of a user electing for specific functions that may use personal data to be accessed by Siri while in a locked state, only if the voice of the user is recognized. This would still make the device available for general non-user-specific Siri queries by other people, without permitting access to data.
In the event the user fails to convince the device of their identity by speech alone, secondary authentication methods are offered, such as other types of biometric security or a passcode.
Apple submits a large number of patent applications every week, but not all ideas make it into commercial products. Considering Apple's existing work concerning biometric security, including Touch ID and Face ID, an extension into voice authentication makes sense, especially when looked at in the context of Siri on the iPhone.
This patent is one of a number Apple has filed in the field, and it certainly isn't the first to surface from the company. For example, a 2011 patent application for "User Profiling for Voice Input Processing" suggests virtually the same idea as the newly-granted patent.
In August, it was revealed Apple was looking into using voice prints for a slightly different purpose: differentiating between multiple users.
Aside from Siri on iPhone, another area this could be useful is with Siri for HomePod, a version of the digital assistant that can only provide verbal responses on hardware without any visual or physical biometric security. Multi-user support has been touted for the smart speaker before, with user voice recognition enabling the possibility of personalized query results or even playing an individual's private playlist.
Granted by the U.S. Patent and Trademark Office on Tuesday, the patent for "Device access using voice authentication" is relatively straightforward, namely detecting the speaker's voice for a vocal request and determining if it is the main registered user.
According to the patent, a device capable to receive speech input from a user could conceivably create a "voice print" for the owner, using multiple examples of their speech. This text-independent voice print would determine characteristics of the user's voice to create a model, either on the device or via an external service, with the result being the point of comparison for future checks.
The voice print could use "statistical models of the characteristics" of a user's pronunciation of phonemes, to create the signature. In theory, this could include characteristics such as voicing, silences, stop bursts, nasal or liquid interference, frication, and other elements.
By being text-independent, the voice print would theoretically be able to determine the speaker's identity just from normal speech, without requiring the use of a specific passphrase. This would be a better way of dealing with voice-based authentication in public situations, where users may not wish to loudly state a passphrase they would rather keep private.
At the time the device receives a vocal command, it effectively creates a model based on the uttered phrase, which it then compares against the already-established voice print. If there is a certain threshold met for similarity, the device could be unlocked then carry out the spoken command.
It is also suggested a voice print match may not need to unlock the phone entirely, but could still enable the spoken request to be performed for the user. For example, Siri on a device could provide a verbal and text-based response to a vocal query about a general topic if it detects the user's voice, but would refrain from providing a response involving user data unless the device is unlocked by other means beforehand.
There is even the possibility of a user electing for specific functions that may use personal data to be accessed by Siri while in a locked state, only if the voice of the user is recognized. This would still make the device available for general non-user-specific Siri queries by other people, without permitting access to data.
In the event the user fails to convince the device of their identity by speech alone, secondary authentication methods are offered, such as other types of biometric security or a passcode.
Apple submits a large number of patent applications every week, but not all ideas make it into commercial products. Considering Apple's existing work concerning biometric security, including Touch ID and Face ID, an extension into voice authentication makes sense, especially when looked at in the context of Siri on the iPhone.
This patent is one of a number Apple has filed in the field, and it certainly isn't the first to surface from the company. For example, a 2011 patent application for "User Profiling for Voice Input Processing" suggests virtually the same idea as the newly-granted patent.
In August, it was revealed Apple was looking into using voice prints for a slightly different purpose: differentiating between multiple users.
Aside from Siri on iPhone, another area this could be useful is with Siri for HomePod, a version of the digital assistant that can only provide verbal responses on hardware without any visual or physical biometric security. Multi-user support has been touted for the smart speaker before, with user voice recognition enabling the possibility of personalized query results or even playing an individual's private playlist.
Comments
Otherwise we're going to see a lot of this:
When will the superior privileged brains in Silicon Valley would stop chasing after "cool" stuff which is short of actual practical use. Look at just about any operating system to understand my point, piles of cool stuff which few people know, care about or ever use. Siri a failure and Apple may have know this but they feared being left behind in this fantasy ego race to get AI in all of our hands. Fortunately most people have hands which operate quite well but if Apple wanted to develop Siri for those with disabilities then I would applaud them. For now Siri is a useless failure.
I use Siri a lot for my smart home functions, texting, calling, weather, timers, measurement conversion, basic references, sports queries, music control and Apple TV use.
With the deployment of silicon containing the Neural cores (A11 chips and up), I expect Siri to gain major improvements in processing. In the near future, I don’t see why Siri couldn’t be processed at the device level.
You think Siri is a failure because it doesn’t do web searches well? I use Siri every day for multiple things that have nothing to do with web searches and am more than satisfied with the results. A sample of things I’ve used Siri for in the last 24 hours: turn lights on/off, set the thermostat, request music, math and conversions, send/read texts, ask the location of a friend, request the time in another country and more. Now that Siri can Arm/Disarm my alarm system it’s getting even better.
Those are are all things that happen faster than taking out my phone, finding and opening an app and doing it manually, by the way.
Some store chain made a commercial based on the theme of voice patterns.
https://www.youtube.com/watch?v=sgJLpuprQp8
At the end it says “The simplest is often the best”
Especially with an Apple Watch, it is far, far faster to say "Remind me on November third that [act] is playing at [club] at 9pm" than to dig out the phone, launch the reminder app, and type all that in.
I'd certainly love it if Siri were as smart as Google Assistant about opening and closing times for some businesses, but it works pretty well for a range of daily tasks that used to take more time and effort to accomplish, and Siri has really done a lot to help me get more organised (formerly a major failing of mine).
Regarding what I bolded in your comment, I've mentioned this before, I think this is a problem with Maps and not Siri (obviously it doesn't make Siri look good, at any rate). People were recently complaining about Siri returning a list of locations when they asked for a specific location AND getting the same results when typed into Maps. I think what we're seeing there and when asking for opening/closing times is Siri querying Maps for that info and returning what Maps provides. So, yes, Siri isn't giving the best results but that's only because Maps isn't providing the best results. If Maps' results improve then so will Siri's responses. To me, this is similar to blaming your television for showing you the wrong channel when it's really the cable box that has the problem.
Don't get me wrong, I'm don't think Siri is flawless and I welcome any improvements. I'd like it if Apple brought Siri's functionality a little closer across devices. I get it that TV Siri doesn't need to provide navigation, but it would be nice to occasionally set a timer on it. Apple added multiple timers to HomePod (a feature I won't use particularly often myself but at least it helps to quiet some of the people who would use it) but I still can't set multiple timers on my iPhone, it's just odd. There are a bunch of little things like that that seem relatively easy to fix/adjust, even if it was simply to give less reason for people to complain.