callstats
At first glance, WebRTC and IoT may not seem to have much in common at all, other
than being trendy topics in both industry and academia. After all, the IoT focuses on
things. It connects embedded devices to the Internet to share data they collect and
interact with, and more in general, promote machine-to-machine communications. In
contrast, WebRTC is all about enabling and simplifying communication between people ,
thanks to its emphasis on audiovisual communication via web browsers and mobile
applications.
However, there are actually several areas where the two work together. There is a distinct overlap between connected devices and human interaction, specifically where their function can be improved through person-to-person communication.
There are numerous areas where the intersection between IoT’s device-to-device communications and WebRTC’s human-centric communications creates opportunities for augmented user experience. For example, Amaryllo sells a line of home security products that use sensors to detect when someone is at the door. Rather than simply taking a photo or alerting the homeowner when the sensor is tripped, Amaryllo’s iSensor HD incorporates WebRTC to offer video streaming, allowing the homeowner to see exactly what is going on. Other products like Ring go even further by offering real-time communication that enables the homeowner to speak with their visitor whether they are at home or not.
During the event, we would like to work exactly at the intersection of IoT and WebRTC. Specifically, we want to look in the domain of IoT home security systems that use WebRTC for video interaction.
A single deployment of a home security system involves numerous surveillance cameras that continuously stream video data to service providers. This produces an enormous amount of video content that needs to be analyzed for potential threat detection.
For successful video content analysis, it is critical to ensure good quality of streamed video. Many elements depend on it, including, for example, optical character recognition algorithms. The challenge for analysing video quality in these home security services is that:
- Video is captured continuously, forever.
- The vast majority of video content is not interesting, as homeowners only care about the quality of video content when something important happens.
For any home security service to be successful, the surveillance system should be able to identify objects and items in the video. When the media quality is poor, the detection algorithms are unable to correctly identify and classify events in the video. Hence, the application must be able to identify issues with video quality in real-time and fix them as quickly as possible.
To this end, we want to build a WebRTC-based IoT monitoring system that is able to analyze measurement data coming from a large number of camera devices and identify issues related to poor video quality in real-time.
With this system, we aim to address one of the following problems:
- How can we use the WebRTC Statistics API to identify that a stream suffers from poor video quality?
- How can we reduce the amount of measurement data sent to our monitoring system? Ideally, we want to have enough data to identify poor video quality while at the same time we do not want the measurement data to compete for bandwidth with the video streams.
- What is the smallest set of data that would allow us to detect failing video quality? The three challenges described above are examples of problems that need to be addressed by successful home security products. There are many others beyond this. We aim to make this hackathon challenge open-ended. If a team is able to identify different challenges of similar complexity within this domain, we are open to them working on it.
The challenge is to build a simple WebRTC-based monitoring system for a home monitoring service that analyses video measurement data coming from a large number of camera devices and detects in the real-time quality issues without looking at the actual video content. To build such a system, the team must:
- analyze video measurement data metrics,
- identify subset of metrics that are useful for detecting quality issues,
- design a protocol for monitoring video measurement data, so that it does not compete for bandwidth with video monitoring.
In order for us to complete these challenges, callstats.io provides the following tools:
- Traces of home monitoring video and its corresponding measurement data.
- A WebRTC app for sending monitoring video.
CALLSTATS I/O Oy is a Software as a Service company based out of Helsinki. It provides products that measure and manage the performance of real-time media communication. The Callstats.io product helps software developers set up, build, and scale communication applications quickly. Founded in 2014 and originally known as Nemu Dialogue Systems, Callstats.io was named a Cool Vendor in Unified Communications in 2016 by research firm Gartner. The Callstats.io product integrates with various third party SDKs and PaaS solutions which make it easier to build and deploy WebRTC applications.