Initially funded through crowdfunding, Cocoon's management team has a good track record when it comes to setting up start ups and have concluded a number of successful exits.
One of the founders, Dan Conlon established cloud storage service Humyo, which was sold for $18million to Trend Micro in 2010 and, while still in secondary school, started the web hosting company Donhost and sold that for $11m in 2005.
Sanjay Parekh and the company's other founders – Colin Richardson, Nick Gregory and John Berthels – have all held various roles at Humyo and Trend Micro.
Smart security
While there may not be a shortage of 'smart' security devices in the market at the moment, such devices tend to be relatively simple, combining motion detection with an Internet connected camera and supported by a cloud-based service with smartphone apps that alert a user should it detect an intruder.
Cocoon has developed a single device that uses 'infrasound' to detect movement by listening to things people can't hear.
"With this new security device, we have set about correcting the main problem with existing sensors, where there are way too many false alarms triggered by the suburban sound track," explains Conlon.
"We don't think that current sensor technology is actually fit for purpose. We wanted to improve it, make the system easier to install and make it simpler to manage. Most people in the UK don't have a home alarm system; it's expensive and complicated. Those that do, often forget to set their alarm, worry about false alarms disturbing neighbours or having to pay the police for erroneous call outs.
"We'd all had bad experiences with existing home security technology and wanted to create something that could protect the whole household but which could be plugged in to a wall socket in much the same way you would a kettle," Conlon notes.
Cocoon's device uses infrasound, a subsound technology which is outside the range of normal human hearing.
"Humans hear within the range from 20Hz to 20kHz. Infrasound is classified as sound below that of human hearing," says hardware engineer Nick Gregory. "Cocoon uses a powerful microphone to detect infrasound and this subsound technology combines sound wave detection (audible and infrasound) with digital profiling and machine learning."
According to Gregory, subsound technology enables Cocoon to detect and make sense of the noises created by activity throughout the home so it can notify you if anything unusual happens.
A key benefit of subsound technology is that it enables Cocoon to detect and profile activity throughout the home without the need for additional sensors.
"It means you only need one small Cocoon device in order to keep the typical home or apartment secure, eliminating the need for expensive wired home security systems and you don't have to remember to arm or disarm it," Conlon continues.
Infrasonic sound waves
By 'listening' to infrasonic sound waves, the company claims the device can monitor an entire home, including detecting intruders through closed doors and in rooms other than where Cocoon is situated. This potentially solves the shortcomings currently associated with existing camera/motion detection based smart home security systems, which are either limited to a single room or require the setting up of 'zones' using multiple devices.
"We believe this is a smarter, more accessible way to deliver home security," suggests Conlon. "Not only is the device a lot simpler, but machine learning algorithms also learn what is normal for an individual's home and so help to reduce false alarms. In turn, smartphone alerts deliver actionable information, enabling you to take meaningful action. I'd like to think that we're finally putting people at the heart of home security."
In addition to its infrasonic sound sensor technology, the Cocoon device also contains a traditional motion sensor, microphone and HD camera with night vision and wide angle lens.
"Cocoon incorporates a HD camera, sensitive microphone and PIR sensors. A Freescale i.MX6 SoC is at its heart, with an Aptina image signal processor delivering a high quality video stream from the image sensor," says Gregory.
"The i.MX6 was chosen for its onboard H.264 encoding and AES encryption, as well as for the ARMv7 instruction set, which allowed us to develop its software rapidly in higher level languages such as Golang."
While the Cocoon engineering team is exploring silicon from a wider set of vendors, Gregory explains that, as the company began life as a crowdfunded start-up, it meant that 'Freescale was one of just a few manufacturers that was able to provide us with the necessary support."
Cocoon has been set up to analyse the readings from its sensors continually. A notable audio event can then be processed to extract an audio 'fingerprint'.
"Different kinds of sounds – for example, jangling keys, the click of a door or loud music from next door – all have different fingerprints," explains Conlon. "In this way, Cocoon can turn the raw stream of audio data into a more comprehensible stream of events, with associated data. These events can then be analysed to discover patterns so, as time passes, Cocoon learns what is normal for an individual household."
This data is then analysed and used to build a model of what events are expected, tailored to an individual's specific household.
"With this model and augmented by information from an individual's mobile app about when they are home or away, Cocoon can decide if a sensor event is significant and show it to the user," says Conlon.
Security is an important issue and Cocoon uses the i.MX6, according to Gregory, because it has features such as secure JTAG, high assurance boot and the ARM TrustZone.
"We're using a well maintained Linux distribution and following security best practice, including client certificates for authentication and end to end AES encryption to protect users' data. The software stack has been designed with an active update strategy to facilitate frequent firmware updates without the need for user intervention," Gregory concludes.
The Cocoon device is currently going through engineering and validation testing and is expected to enter production at the end of 2015.
