Patent Grant
7161957
1. Technical Field
The invention pertains to wireless communications and, more particularly to optimally sending video packets over a wireless link under varying bandwidth based on video quality demands.
2. Description of the Related Art
A universal mobile telecommunications network 10 of
Referring to the diagram 20 of
Delivering a variable bit rate (VBR) compressed video stream over a wireless network or link requires network resource allocation. Delivering a variable bit rate (VBR) video stream with variable frame sizes over a wireless mobile network requires appropriate network resource allocation for each frame to maximize the network efficiency, minimize the delivery jitter (frame arrival variation), and smooth the traffic flow over the wireless network.
Current video delivery application maps a video stream into one quality of service QoS queue without any adaptation. For a variable bit rate VBR video stream, a fixed quality of service QoS queue for all frames with variable sizes is not an optimal solution.
Accordingly, there is a need to optimally send variable length video packets over a wireless link.
The present invention optimally sends video packets of variable length or variable bit rate over a wireless communication link.
A computer readable medium containing instructions that, when executed by a processor in a communication network, performs the steps of receiving packets of variable length of compressed video information; and determining a queuing delay and partial bandwidth from a total bandwidth available on a wireless channel for each of the packets for transmitting the packets over the wireless channel, the queuing delay and partial bandwidth for each of the packets being based on criteria for reception quality at a receiving location on the wireless channel. Preferably, the partial bandwidth is one of a plurality of distinct queuing delay and partial bandwidths representative of wireless transmission quality of service classes and the plurality of distinct queuing delay and partial bandwidth are representative of four quality of service classes: conversational, streaming, interactive and background. Preferably, the criteria includes at least one of a maximum bandwidth (kbps), a guaranteed bit rate (kbps), a delivery order, a maximum service data unit SDU size (octets), service data unit SDU format information (bits), a service data unit SDU error ratio, a residual bit error ratio, a delivery of erroneous service data units SDUs, a transfer delay (ms), a traffic handling priority, and an allocation/retention priority.
In another aspect of the invention, an apparatus in a wireless network includes a video encoder for encoding uncompressed video information into packets with variable length of compressed frames, and a decision processor for selecting an appropriate QoS class associated with a partial bandwidth on a wireless channel for each of the packets based on an intended reception quality.
In a further aspect of the invention, a method for optimal bandwidth allocation for wireless transmission of video information, the method comprising the steps of converting video information into packets with variable length of bits; and allocating a bandwidth on a wireless channel for each one of the packets, the bandwidth being associated with criteria for reception quality of the packets over the wireless channel.
These and other objects, features and advantages of the present invention will become apparent from the following detailed description of illustrative embodiments thereof, which is to be read in connection with the accompanying drawings.
This disclosure will present in detail the following description of preferred embodiments with reference to the following figures wherein:
This invention addresses the mapping of a variable bit rate VBR video stream into four Quality of service (QoS) classes specified in the 3GPP's Universal Mobile Telecommunications System (UMTS) specification in order to maximize the network efficiency, minimize the delivery jitter (frame arrival variation), and smooth the traffic flow over the wireless network. The mapping is determined by optimizing an objective function, which achieves the frame constant delay objective. By mapping frames with a large frame size onto a higher quality of service QoS class in a wireless network and the frames with a smaller frames size on a lower quality of service QoS class provides a simple state-less queuing scheduling mechanism to balance load cross quality of service QoS classes to achieve a high call admission rate and low video quality distortion. In addition, traffic carried in lower quality of service QoS class costs less than that in higher quality of service QoS class. This invention provides users with a balanced cost versus a quality scheme to tailor users requirements and expectations of a video delivery application over a wireless network.
Typically, video frame types have the following order in frame sizes: I, P, B. Usually, the frame size of an I frame is much greater than that of the P and B frames. Therefore, this invention spontaneously maps I frames into a higher quality of service QoS class and P and B frames into a lower quality of service QoS class. From a scalable video compression point of view, I frames can be considered as a base sublayer with highest priority, and P and B frames as enhancement sublayers with lower priority. Therefore, this invention automatically implements the idea of mapping a scalable video into different quality of service QoS classes.
An inventive wireless communication system 30 with multiple quality of service QoS classes Q0, Q1, Q2, Q3 is shown in
In a 3G mobile network, four quality of service QoS classes are specified: Conversational, Streaming, Interactive and Background. By mapping video frames with different sizes into more than one mobile network quality of service QoS classes, we are able to achieve those goals.
In a video, not all the frames are created equal. Some frames are more dynamic than the others. For example, a sport video with lots of motions and video with a scene change in the middle of a video recording require much more number of bits to encode and more bandwidth to deliver. A video is digitally compressed into a series of frames with variable sizes (number of bytes) to maximize the overall video quality. Using a network with fixed Quality of service QoS (throughput, delay time, jitter requirements) is not economical. In a 3G universal mobile telecommunication system UMTS wireless network, 3GPP standard body specified four different quality of service QoS classes: Conversational (quality of service QoS class 0), Streaming (quality of service QoS class 1), Interactive (quality of service QoS class 2) and Background (quality of service QoS class 3). Each class has its own range of quality of service QoS parameters. 3GPP specifies the following parameters for those four quality of service QoS classes: maximum bandwidth (kbps), guaranteed bit rate (kbps), delivery order, maximum service data unit SDU size (octets), SDU format information (bits), SDU error ratio, residual bit error ratio, delivery of erroneous SDUs, transfer delay (ms), traffic handling priority, allocation/retention priority.
Suppose a video stream consists of a series of frames,
F={f0, f1, . . . , fn}
and each frame has the size in number of bits,
B={b0, b1, . . . , bn}
In a universal mobile telecommunication UMTS network 10, four data queues are allocated to support four quality of service QoS classes for each user. At time t, we assume the estimated available bandwidth for each queue is available:
C(t)={c0(t),c1(t),c2(t),c3(t)}
and the average queuing delay time for each queue is known as well:
D(t)={d0(t),d1(t),d2(t),d3(t)}
At the same time t, a frame fi(i=0, 1, 2, . . . ,n) with frame size bi (i=0, 1, 2, . . . ,n) moves to the beginning of the transmission buffer. The time it takes to transmit the frame from the transmitter to a receiver with one hop through one of the four quality of service QoS queue is
Tki=dk(t)+bi/ck(t) i=0, 1, . . . , n k=0,1,2,3→ Equation (1)
The first term is the queuing delay and the second term is network transmission delay. The end-to-end delay requirement, ΔT, of a video stream is one of the video qualities the underlying network must guarantee. The variation of the delay from frame-to-frame is known as jitter. A video stream delivered without jitter (jitter=0) meets the constant delay requirement, i.e.
ΔT=const
However, jitter is inevitable in a video delivery network. Therefore, we define the following objective function and optimize it in order to minimize the jitter, and therefore to keep the delay time as constant as possible.
where α and β are constants and determine the range of jitter. This constraint says the delivery time cannot be earlier than αsec and later than βsec using ΔT as a reference. If every frame must be delivered within ΔT sec, β must be equal to zero. This is illustrated in the graph 40 of
In other words, we search for all the four quality of service QoS classes to find the one with the lowest available bandwidth quality of service QoS that still meets the delay requirement.
Here is an example to illustrate the algorithm proposed above. Suppose the ith frame is transmitted such that the end-to-end delay constraint ΔT is 10 ms. Let the time it takes to transmit the frame from the transmitter to a receiver with one hop through the four Quality of service QoS queues is
Tki=dk(t)+bi/ck(t) i=0, 1, . . . , n k=0, 1, 2, 3
and using Equation 2
Using Equation 3 giving the constraint that α≦Ω≦β, where α=−1 ms and β=2 ms, we see that Quality of service QoS 2 should be chosen to send the frame as per the above algorithm to achieve the best cost-effective objective even QoS 0 and QoS 1 provides higher bandwidth.
The inventive of mapping a variable bit rate VBR video stream into the four quality of service QoS classes has three major advantages over a traditional approach: It optimizes the utilization of the quality of service QoS resources in a UMTS system in an optimized efficient manner. Using a lower permissible bandwidth quality of service QoS class has economical benefit for both operators and users.
The objective function is optimized (minimized) for jitter variation. Near constant delay requirement can be achieved. By achieving the near constant delay requirement is equivalent to a traffic shaping function from the network point of view. This reduces the burst characteristics of delivering a video stream over a network.
Having described an exemplary mobile telecommunications system employing the inventive optimal transmission of variable length packets over a wireless link (which is intended to be illustrative and not limiting), it is noted that modifications and variations can be made by persons skilled in the art in light of the above teachings, for instance, a Hiperlan II wireless LAN (WLAN) with QoS provisioning. It is therefore to be understood that changes may be made in the particular embodiment of the invention disclosed which are within the scope and spirit of the invention as outlined by the appended claims.
| Number | Name | Date | Kind |
|---|---|---|---|
| 6269078 | Lakshman et al. | Jul 2001 | B1 |
| 6967921 | Levy et al. | Nov 2005 | B1 |
| 20020054578 | Zhang et al. | May 2002 | A1 |
| 20020114292 | Kawabata et al. | Aug 2002 | A1 |
| 20020154656 | Kitchin | Oct 2002 | A1 |
| 20030235196 | Balachandran et al. | Dec 2003 | A1 |
| 20030235208 | Sahinoglu | Dec 2003 | A1 |
| 20030236904 | Walpole et al. | Dec 2003 | A1 |
| 20040105393 | Ronneke et al. | Jun 2004 | A1 |
| Number | Date | Country | |
|---|---|---|---|
| 20040156354 A1 | Aug 2004 | US |
