Packet-Switched Streaming Service in Non-Bitrate-Guaranteed Mobile Networks
Research output: Book/Report › Doctoral thesis › Collection of Articles
|Publisher||Tampere University of Technology|
|Number of pages||87|
|Publication status||Published - 3 Jun 2011|
|Publication type||G5 Doctoral dissertation (article)|
|Name||Tampere University of Technology. Publication|
|Publisher||Tampere University of Technology|
Streaming allows consuming content, such as video or audio, without storing it on a receiving device. The benefit of streaming is that it requires little memory from the device. The drawback is that the device must always be connected to the network and that the network connection must offer adequate capacity and performance stability. These network connection requirements pose a number of challenges especially for mobile devices connected to packet-switched networks, in which the basic network connections cannot guarantee the allocated bandwidth or limit delays under certain levels.
A comprehensive study and a large set of simulation tests were carried out to analyze the characteristics of packet- switched networks that cannot guarantee the capacity. In this dissertation, such networks are referred to as “non-bitrate-guaranteed mobile networks”. This study revealed several key problems including: data transfer gaps caused by cell reselection, bandwidth fluctuation caused by either signal quality or cell load, streaming client buffer underflow and limited capability of the streaming client to report these problems to the streaming server. Novel solutions to these problems are proposed in this thesis. In addition to these service issues, this thesis evaluated subjective quality thresholds of lip synchronization and video. It is equally important, compared to technical improvements, to understand whether the improvement give subjective benefits. For example, increased bitrate improves the video/audio quality of streaming service, but the device hardware may become limiting factor that improvements are no longer noticed.
Four fundamental problems were investigated in the thesis. The first problem is related to cell reselection. This is a natural phenomenon of every mobile network, but unfortunately, it has a tendency to cause gaps in the dataflow, which causes noticeable quality degradations in media streaming. This thesis provides a novel solution to hide the cell reselection from users and proposes a novel mechanism to maintain a full streaming client buffer. The proposed solution makes use of standard Packet-switched Streaming Service (PSS) methods and requires no changes to the network.
The second fundamental problem this thesis examines is related to how network capacity may change if the mobile device moves into a position that changes the received signal quality. Another major factor affecting network capacity is the amount of devices served by the network cell. The change in capacity can be positive or negative. Streaming service users are unable to detect positive capacity change from the streaming client, which simply means that the network could offer better service than what is currently being provided. Negative capacity change, on the other hand, is easily detectable by the user. When the streaming client buffer underflows, the quality drops drastically. The thesis provides a novel solution for the streaming client to report network capacity changes to the streaming server. Again, an additional merit of the proposed solution is that it uses standard Packet- switched Streaming Service (PSS) methods and requires no changes to the network.
The range of options for reporting streaming service problems or simple statistical characteristics is very limited. Real-time Transfer Control Protocol (RTCP) offers a limited set of predefined parameters that allow reporting problems. This is the third fundamental problem investigated in the thesis and a novel solution for reporting problems and statistical information from client to server is proposed. The method allows the client to provide raw data to the streaming server, which may include relevant statistics about the quality of the service. It also offers a parameter negotiation technique allowing the client to easily expand the current set of parameters used and a better adaptation to different types of client devices.
When improving streaming services in mobile devices, it is important to ensure that such improvements are noticeable by the user. Due to the limited size of a mobile device display, improvements in content streaming may not be visible and prior measurements in e.g. the television industry may not apply. This fourth problem is investigated in this thesis and new parameters specific to mobile devices are determined. The first objective is to determine the bounds within which synchronization of audio and video (i.e. lip synchronization) may vary and still produce acceptable results for the user. The second objective is to determine the correct ranges of bitrate and frame rates that should be used by service providers in order to guarantee a certain streaming content quality. A comprehensive study was carried out and suitable parameters and ranges are proposed for mobile streaming applications.
The main emphasis in this study is on the Non-bitrate-guaranteed networks, which illustrate the basic and most limited mobile networks existing at the time the research was carried out. Most of these issues mentioned above have already been solved in more advanced networks, but such networks have limited abilities to solve such problems for all users, since it may be too costly to offer them to all users. The elegant solutions described herein require no changes to the network; therefore, availability would not be limited by the network. If the service provider chooses to implement these in mobile client and service servers, they would be available for all users regardless of the underlying network.