March 29, 2018

Knock knock. Who’s there?

My colleague Chris recently wrote about the growth in front-door tech and we’ve been looking closely at this area of the smart home, looking at how sound recognition enhances existing consumer benefits, as well as enabling new opportunities or overcoming potential product weaknesses.

Video doorbells are becoming big business. Nest have launched Hello in the US, market-leader Ring has been acquired by Amazon, who themselves are developing their Key service linked to Prime. The main marketed benefits of these products are principally around:

  • Interacting with visitors (desired and undesired – friends, neighbours, delivery drivers, time-wasters, salespeople or potential sneak thieves)
  • Keeping an eye on packages
  • General home security.

In all of these cases, the two user interaction points are person detection (motion and facial recognition) or the doorbell being pressed. When both of these interaction points happen, then the doorbell alerts you to movement and then tells you that somebody has rung your bell. Which is great unless either or both of these interaction points aren’t triggered. Then the doorbell’s effectiveness is diminished.

Let’s look at person detection to start with. As with all motion-based video analytics platforms, consumers face a decision on their comfort level with false alerts. How many will they get and how willing are they to respond each time? Depending on where your front door is, with regards to the road and path, there may be a number of false alerts. If there are too many then consumers will turn off motion detection because of the irritation factor. It is an issue with connected devices, which is why accuracy is so important (ie, intelligent sound recognition v noise detection).

Without motion detection turned on, a consumer is unaware of a potential threat (unless they ring the doorbell).

And this brings me to the other critical interaction point: the doorbell itself. If the visitor knocks on the door and doesn’t press the doorbell (intentionally or unintentionally) then the consumer can’t detect whether somebody seeing if you are home. Even with motion detection turned on, how can you distinguish between a harmless leaflet distributor or a thief if you don’t know if they have tried to see if you are in? It’s bad enough coming home to a letterbox full of junk mail, who wants to buy a product that amplifies that spam through your phone while you are at work?

By adding sound recognition into the smart doorbell then there are a number of situations that can be overcome. For example:

Delivery driver

The consumer has turned off motion detection because of high levels of false alerts and the driver doesn’t notice the doorbell or prefers to knock. With sound recognition built in, it can respond to the sound of a door knock (and not anything else). This means that the consumer gets alerted via their doorbell and they can interact in the usual way, asking the driver to leave the package, deposit it inside or put it in a safe place. Without the ability to detect the sound of a knock, then the driver either leaves with the package or leaves it outside and the consumer is unaware.

Doorstop theft and breaking and entering

Whether the sneak thief is attempting to steal a doorstep package or looking to break-in, the standard modus operandi is to check whether inhabitants are in or not, before running off with the package or attempting to get into the property. If motion detection is off and if the potential thief avoids the doorbell (highly likely) and knocks instead, then the doorbell can be activated through this sound being recognised, enabling the home-owner to challenge the potential thief. The door knock sound combined with the sound of a dog barking can also provide additional security benefits that motion detection alone may not capture.

Elderly care

In addition to situations around unoccupied homes, there is potential to support elderly care by monitoring and recording doorstop salespeople or distraction burglars. In addition to motion detection and doorbell/knocking, being able to detect human conversation (two people talking) or speech detection (single person talking) then there are opportunities to record these interactions or have family members alerted.

Motion detection and facial recognition are two critically important AI technologies when it comes to smart doorbells but they are not suitable in all situations and may not be used if they fail to meet consumer expectation. Additionally, if nobody presses the doorbell then the consumer misses something really important – somebody is outside their house. This is where intelligent sound recognition has a clear role to play. Simply detecting loud noises is not useful and is highly infuriating for consumers. Instead, these devices should and can (thanks to us) be taught to understand important sounds that can be linked to an interaction.

The result is an intelligent product that does even more to provide the consumer with peace of mind.

>