Issues with real-time communications that can be solved with artificial intelligence are divided into two categories: service level and infrastructure level. Infrastructure level applications are typically developer-facing and optimization-based, while service level applications are mainly customer-facing and feature-based. No one side - service level or infrastructure level - is more important or influential than the other, and both give unique and intriguing improvements to real-time communications.
Infrastructure level applications are implemented primarily on the developer-facing side. This includes optimizations such as audio and video quality, network quality, and improving the quality of experience. These applications can significantly improve the quality of real-time communications, especially with regards to audio and video quality.
What are Some Infrastructure Level Artificial Intelligence Applications in Real-time Communications?
These are but a few of the key infrastructure level applications being developed today.
- Audio and Video Quality: In order to improve video quality in conferences, artificial intelligence is being developed to auto-regenerate image and audio. This class of algorithms generates higher quality images from lower quality images. For example, RAISR: Rapid and Accurate Image Super-Resolution, is an artificial intelligence technique that uses machine learning to create high-quality versions of low-resolution images.
- Optimal Quality: Optimal quality of a conference can be determined through artificial intelligence by assessing network conditions. By altering the optimal quality through choice of codec, internet path, audio-only versus video, resolution, and error-resilience, the best possible user experience can be obtained. At callstats.io, we are actively working on this with Optimize.
- Network Performance: Artificial intelligence can be used to monitor the network more effectively and quickly than a human is able to. Companies like Mist are leveraging this to identify network performance issues.
- Network Resources: Predicting network resource requirements by hand can be very difficult, but artificial intelligence can be used to forecast based on previous trends. Companies like Bellintegrator use artificial intelligence to anticipate network resource usage and allocate accordingly.
- Network Routing: High round-trip time during a video conference can have an adverse effect on media quality and is affected by a number of factors. Artificial intelligence can be used to choose the appropriate media server based on a number of metrics for a specific call.
- Quality of Experience: Assessing Quality of Experience within a call is difficult, since a lot of the factors rely on the perception of the user of an application. Artificial intelligence can be used to predict this quality based on a number of key metrics.
- Congestion Control: Congestion control is a series of techniques used by networks to prevent collapse. Liege University has developed artificial intelligence techniques to improve congestion control over wireless networks.
- Self-driving Networks: Self-driving networks automate functions in the network that were previously done manually using artificial intelligence and historical trend data. Companies like Juniper Networks are building autonomous networks to self-configure, monitor, and manage themselves.
There are of course more infrastructure level applications in development today. These are but a few important ones changing the real-time communications industry.