CHAPTER 2 BACKGROUND AND RELATED RESEARCHES
2.6 Literature Review Summary
Study Adaptation Index Method Parameter Consideration
Ki-jong et al.
[52] (2008) Media access delay
Monitoring module in AP works crosslayerly with a
codec controller module in stations
Codec and packetization
interval (Only G.729.1
bit-rates) Myakotnykh
[85], [34]
(2008, 2009)
E-model Instant quality Integral quality
Adapting packet size then codec (depends on the
situation)
Codec and packetization
interval
Tüysüz et al.
(2010)[35]
MAC layer info RTCP Quality factors
R-factor Capacity estimation
Categorizing PL and switching to proper packet size
and codec switching
Codec and packetization
interval
Tüysüz et al.
(2010) [76]
MAC layer info RTCP Quality factors
R-factor Adaptive jitter buffer
Categorizing PL and switching to proper packet size
and codec switching
Codec and packetization
interval
Qiao
(2004)[79] Objective MOS
Coding rate adaptation Higher priority packets pass during
congestion
Variable bitrate codec and Priority marking
in the user level
Mazurczyk (2007) [82]
Watermarking MOS scale
Adaptation of speech codec (bitrate-frame size) And Adaptation of
the playout buffer size and amount of
data used for Forward Error
Correction
Coding rate Playout buffer
size FEC data length
+ security
Liu et al.
(2011) [84] PLR using SNMP Different type of priority queues
Different categories of
congestion
2.7 Chapter Summary
This chapter has presented the background of adaptive rate VoIP studies. These studies have been classified into four categories based on their adaptation mechanism.
In each category the most related studies have been discussed and their parameters and mechanisms investigated in detail to verify their advantages and disadvantages.
Based on the classification, some of the studies have considered only codec adaptation while some others considered only packet size adaptation. The proposed algorithm in this study will consider both approaches to exploit their benefits.
One of the main disadvantages of the algorithms that used different codec as their adaptation method is high cost including paying a license fee for each codec and/or upgrading the current hardware such as routers or gateways with the new codecs.
Another disadvantage is that switching from the higher bit-rate codec to the lower bit- rate codec has the obvious effect in speech quality perceived by end users.
The algorithms that have studied adaptive packet size as their adaptation method have lower costs in term of paying extra fees and also the modification of the current equipment are not required. They are more transparent to the end user since the codec does not change [49]. Furthermore, packet adaptation can adapt the output rate while codec is fixed. Hence, it has the benefit of using high bit-rate codecs and at the same time adapts the output rate. However, most of codecs support 10-40 ms audio in each RTP packet thereby the level of rate adaptation using packet size is restricted.
Besides, the packetization delay that is posed by the larger packet size limits this approach for the severe congestion, where delay is high and larger packet size adds extra delay.
The basic idea gathered from the above discussion is that both approaches have some advantages and disadvantages. In order to minimize the disadvantages of codec adaptation (high cost) packet size adaptation is proposed for low to, moderate congestion and codec adaptation will be done only in severe congestion where packet size adaptation is fail to reduce the output rate remarkably.
As mentioned above our proposed algorithm considers codec and packet size adaptation to obtain the benefits of both approaches and reduce the drawbacks of them. Yet, in the area of research with consideration of both parameters, there are some researches that have focused only on LAN networks or they have not focused on multi-rate effect that causes by the IEEE 802.11 standard. Furthermore, in some others, their adaptation method needs to be improved to reduce the adaptation cost and/or their calculation parts need to be minimized and/or the right adaptation time need to be determined.
This study tries to address the issues found in previous studies. This research focuses on the multi-rate effect which caused by LA in IEEE 802.11WLANs.
Additionally, in the adaptation process this research tries to find the balance between codec and packet size adaptation. Furthermore, for finding the right adaptation time the focus of this study is on RTCP-XR that is including the set of VoIP quality metrics and network condition information. This research is the foremost work that uses RTCP-XR in its algorithm to reduce the calculation part of previous researches and to get more information that is accurate for adaptation process.
In this study, the method of finding the right adaptation time is to monitor VoIP stream beside delay jitter (as an instant quality metric) and MOS (as result of the general quality metric) which are the best parameters among all the parameters chosen by earlier researches. This method is similar to [26] and [35] but there are two main problems in these two studies. First, they used older RTCP and they need to calculate the R-factor. However, in RTCP-XR VoIP metrics block, R is provided, so calculation part can be omitted which can affect the faster execution of algorithm.
Second, when the transmission rate falls to the lower rate, previous algorithms trigger the codec adaptation immediately, while it is not necessary to adapt the codec in every transmission rate changes, and sometimes the system can sustain these changes which will be investigated later.
CHAPTER 3