Video reception on the move can be improved significantly by combining Scalable Video Coding (SVC) with Quality of Experience Monitoring (QoE) to control playback in the device.
This conclusion has been reached by a French project called SVC4QoE that finished its final phase in April 2012. It could help solve the problem of reliable, high-quality video distribution to wireless devices such as PCs, tablets and smart phones in trains and cars.
SVC is a part of streaming standards such as Apple HLS, Microsoft Smooth Streaming, and the emerging universal MPEG DASH, and so is already being used for delivery of video across fixed networks. It encodes video at different quality levels to cater for varying network conditions and end device capabilities. Quality can vary in three ways — frame rate, spatial resolution and level of detail within each frame.
Compression methods such as MPEG4 operate by attempting to preserve as much picture detail as possible while reducing the number of bits required to store a given number of frames at a particular spatial resolution, but inevitably some quality is lost. The amount of data per second of video and therefore the required transmission bit data rate can be further reduced by cutting either the frame rate or the spatial resolution or both. With SVC, a video stream might then be encoded as a base layer at 15fps, a resolution of 320 x 240, and compressed to yield a data rate of 300kb/s, suitable for mobile reception on a Smartphone.
Additional layers could expand that stream to 720p video at 3Mb/s adequate for a broadcast service to set top box, with convenient stopping points for relatively high quality streaming over the Internet, for example at 640 x 480, 30fps and 720p at 2Mb/s. All these layers are incorporated in a single file, which has two benefits — it reduces the administrative complexity of sending separate streams, and means that at any time a given device can display at the highest quality possible.
On a fixed network, the maximum quality possible can be calculated by measuring the network bandwidth available, assuming the device capability is known. But in the case of mobiles, it is not so easy because quality can fall off a cliff suddenly, through problems such as signal fading and multipath propagation, and these are amplified when the device is moving. The signal path is then constantly changing, causing rapidly varying propagation conditions.
But, the use of a QoE monitoring probe enables the quality to be detected very quickly so that the consumer’s device can respond by changing which layer to play from the QoE stream. If the signal deteriorates suddenly such that the current quality cannot be sustained, it will drop back to a lower or base level. The principle is just the same as fixed line streaming but adapted for the vagaries of mobile transmission.
The SVC4QoE project began in 2009, and in its final phase, field trials were conducted in the Rennes area over an experimental DVB-T2 network, validating the theoretical concepts and gathering operational data.
A key point is that the lowest base level within the stream is sent as securely as possible with extra error correction built in to make sure at least that is always available. The project then demonstrated how when receiving conditions deteriorate, the QoE monitoring probe prevents total “black screen” failure by supplying the SVC decoder with this securely-received base layer.
The project’s transmission chain comprised an off-line SVC encoder and a DVB-T2 gateway, using a signal emitted from two transmitter sites, allowing the testing of both MFN (Multi Frequency Network), and SFN (Single Frequency Network) operation.
The broadcast signals were received in a mobile van where the DVB-T2 demodulator supplied a splitter connected to an SVC decoder. The QoE (Quality of Experience) measurement controlled the SVC decoder to ensure the display of the correctly decoded SVC layers.