Electrical computers and digital data processing systems: input/ – Input/output data processing – Input/output data buffering
Reexamination Certificate
2001-04-13
2004-01-27
Perveen, Rehana (Department: 2182)
Electrical computers and digital data processing systems: input/
Input/output data processing
Input/output data buffering
C710S033000, C710S035000, C710S029000, C710S030000, C710S056000, C711S170000
Reexamination Certificate
active
06684273
ABSTRACT:
BACKGROUND OF THE INVENTION
The present invention relates to a method of buffering a data stream at its reception as set forth in the preamble of claim 1.
In a circuit switching network as actual telephony network, every connection is set-up by blocking a single channel or fixing some bandwidth per connection. Such channel will be release as soon as the users decide to terminate the call. Due to the reservation of bandwidth during the complete call, the quality of the call is predictable. But this type of network has also disadvantages. At first, every connection must be set-up which consumes time and resources. Secondly, every connection has a fixed bandwidth allocated (usually 64 kbps). A user can not ask for more or less bandwidth. Even during silence periods, the resources are used completely. Further unused circuit are unused capacity. Circuit switching does not provide easily the capability to fill this unused capacity with elastic traffic.
All these explains, among others, the interest to use internet protocol (IP) networks to perform usual telecommunications connections. There, no connection must be set-up. Such networks are therefore called connection-less or packet-switched networks. No resources are allocated and connectivity is already provided by IP. In that case the data (audio like voice or video messages) are transferred over that network as data stream made of packets encapsulated into frames. The packets will mostly follow different routes through the network, what will result in different not necessarily predictable delays (jitter delay) between them. Furthermore, the time taken to insert these packets into a network (time access) will vary a lot according to how busy this network will be. This is in fact the main cause of jitter delay. It may even vary substantially during a same telecommunications. This delay is intrinsic to the technology used like e.g. in a LAN ethernet the Carrier Sense Multiple Access/Collision Detect.
Voice over IP (VoIP) is a typical application using such network and requesting a high level of quality of service (QoS). A real time communications between a sender and a receiver will then be possible only if such requirement of QoS will be fulfilled. Therefore, it is crucial to control and minimize all delays occurring on such a network. This is obtain among other by storing at the receiver some amount of received data in a buffer before restituting it (timestamp-dejittering). In such a way it is possible to rebuild the data stream with the same sequence of packets as when the data stream left the sender.
The delays occurring on a packet-switched network are mostly of two nature, one fixed due to the speed of the stream on that network, and the other variable (jitter) due to the chosen path and mostly to the amount of the already present traffic in that network. Since latter can change a lot from a communication to another or even during a communication e.g. due to some overload of the network at some special hours of the day, it is of interest to adapt some characteristics of the used buffer to the circumstances.
In WO97/39409 is disclosed a communications system utilizing a buffer of variable size. That invention is developed in the field of native processing environment modems such to permit to adapt the software controlling the modem to the constraints of changing standards (e.g. V.32 and V.34). Two different buffer sizes are available. The choice is made according to the latency time period desired. There, small buffers will provide the communication system with short and accurate response times. On the other hand, increasing the buffer size will make the modem processing more robust, reducing the implementation overhead, making it more resistant to operating system latencies and giving it greater flexibility in the exact time within the time slice i.e. between interruptions occurring during processing without necessarily disturbing the transferred data stream.
The available sizes for the buffer are already predefined in WO97/39409. Therefore, the choice of the buffer size is very restricted in fact to only two values. Furthermore, the switching from one buffer size to the other one is performed in a fixed manner at some definite time identical for all communications. This method is absolutely not adequate if high QoS is demanded for a transfer of data stream on a packet-switched network where the delay (jitter) of each packet sent through it may be different and not predictable.
SUMMARY OF THE INVENTION
It is an object of the present invention to optimize the reception quality in a connection-less network such that real time telecommunications with a very high QoS are possible.
This object is attained, according to the invention, by a method of buffering a data stream at its reception as claimed in claim 1.
To overcome the problem of the non-predictable delay taken by the packets when inserted into a packet-switched network and transferred over one almost randomly chosen path of that network, it is of great advantage to store temporarily part of the data stream at its reception on the receiver. This is performed using a buffer of specific size where the packets will be store and classified in the right order. Indeed, the delay taken by different packets can be so different that some packets may reach the receiver before other packets sent first by the sender.
The size of that buffer must be chosen such that it is big enough to permit to rebuild at the receiver the same data stream as the one leaving the sender. On the other hand, this buffer size must be kept as small as possible so not to provoke any audible scrambling or chopping. Advantageously, the method according to the invention comprises the steps of measuring regularly the delay (jitter) of at least one packet when reaching the receiver. Afterwards, said measured delay is compared in some way with a predefined delay obtained e.g. by some averaged estimation of the time taken by several packets for the transfer on that network. And finally, the result of that comparison is used to adapt regularly the buffer size. If, for example, the measured delay is much bigger than the estimated one, it will be a hint that with such to small buffer some packets will get lost. Therefore, the method according to the present invention comprises the step of taking an action onto the size of the buffer i.e. in that case the possibility to increase that buffer size. On the other hand, if the measured delay is much smaller than the estimated one, it will be a hint that the buffer size has to be reduced increasing then also the QoS of the performed telecommunications over such packet-switched network.
An embodiment of said method of buffering according to the present invention is characterized in that the action on the buffer size is taking following some predefined criteria. Latter consists of several different cases to which the result of the performed comparison shall be affected according to its value, and which will imply a change or not of the buffer size. Such step is advantageously applied regularly so to provide a dynamic adaptation (auto-adaptive) of the buffer size to the present circumstances of the network. The different sizes of the buffer are not predefined but rather completely variable and limited at last by the unity used for storing said data into said buffer.
REFERENCES:
patent: 4604717 (1986-08-01), Kaplan
patent: 5652627 (1997-07-01), Allen
patent: 5916309 (1999-06-01), Brown et al.
patent: 5974518 (1999-10-01), Nogradi
patent: 5996031 (1999-11-01), Lim et al.
patent: 6360271 (2002-03-01), Schuster et al.
patent: 6385673 (2002-05-01), DeMoney
patent: 6438604 (2002-08-01), Kuver et al.
patent: 6487196 (2002-11-01), Verthein et al.
patent: 6487603 (2002-11-01), Schuster et al.
patent: 35715912 (1982-10-01), None
patent: WO 97/39409 (1997-10-01), None
Boulandet Patrick
Burger Philippe
Querelle Rodolphe
Zaun Jean-Marc
Alcatel
Perveen Rehana
LandOfFree
Auto-adaptive jitter buffer method for data stream involves... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Auto-adaptive jitter buffer method for data stream involves..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Auto-adaptive jitter buffer method for data stream involves... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3218905