BOARD OF STUDIES FOR INFORMATION TECHNOLOGY
Department of Computer Science and AI
B.Sc. I.T. (Hons.) Year IV
June 2000 Examination Session
CSA402: Multimedia Systems (2 credits) 6th June, 2000
900-1130
___________________________________________________________
This paper contains six questions, you are to attempt four.
___________________________________________________________
Question 1 (50 marks)
1a Assume that 1 (one) live video stream and 2 (two) live audio streams (one for the left channel and the other for the right channel) are to be delivered from the same server to the same remote sink using asynchronous data transmission. The three streams relate to the same event (e.g., live broadcasting of an operatic event), but are individually captured and compressed using dedicated analogue to digital convertors. Should the streams be synchronised and combined at the source or at the sink? Justify your answer.
[30 marks]
1b What would your answer be if the streams were transmitted in isochronous transmission mode?
[10 marks]
1c What would your answer be in the event that only one of the media streams was continuous and the other two were discrete?
[10 marks]
Question 2 (50 marks)
2a Briefly explain the process of resource reservation and management in a typical multimedia system.
[15 marks]
2b Use the Continuous Media Resource Model to calculate the maximum average data rate (M), maximum burstiness (B), and maximum buffer size (S) required for voice-quality telephony over a 10Mbps LAN, given that the audio signal is sampled at 8 kHz and coded with 8 bits. In your calculations, assume that a network packet is at most 12000 bytes, and that voice-quality audio data is transmitted at a rate of 40 messages per second.
[25 marks]
2c Describe what could happen to the application performance and level of network traffic in the event that each parameter (M, B, S) is under- or over-stated.
[10 marks]
Question 3 (50 marks)
3a Explain the various steps during which synchronization must be maintained to deliver live multimedia data in a distributed environment. Use examples liberally.
[15 marks]
3b Explain the quantization step in general compression and decompression techniques. Give examples of when quantization is used in the compression of still images, video and audio.
[15 marks]
3c Explain JPEG's lossless mode of operation. What compression ratio is typically achieved?
[10 marks]
3d Explain, using examples, the difference between periodic and irregular media tasks, and the advantages and disadvantages each poses to a real-time process scheduler.
[10 marks]
Question 4 (50 marks)
4 In information retrieval, two users with different requirements may end up expressing identical queries. An information retrieval system will present both users with identical sets of documents deemed to be relevant. Discuss the advantages and disadvantages of this approach, and suggest two ways (one adaptive and one non-adaptive) in which users can be given interactive or automatic assistance to reformulate their query to better describe their information requirement.
Question 5 (50 marks)
5a Explain the motivation for the statistical model of information retrieval.
[15 marks]
5b Give the cosine similarity measure for the statistical model of information retrieval, and explain how it supports the motivation for the model.
[15 marks]
5c Given the query Q = (1, 0, 1, 0, 0, 1), use the cosine similarity measure to rank the following documents.
D1 = (1, 1, 0, 1, 1, 1)
D2 = (1, 1, 1, 1, 1, 1)
D3 = (1, 0, 0, 1, 1, 0)
D4 = (0, 0, 1, 0, 0, 1)
D5 = (0, 0, 0, 0, 0, 1)
D6 = (0, 1, 1, 1, 1, 0)
[12 marks]
5d Explain how weights are assigned to terms in the statistical model of information retrieval.
[8 marks]
Question 6 (50 marks)
6 Information can be used for retrieval only if it has been extracted from the source and indexed. In full-text information retrieval systems a typical solution is to extract all content-bearing words from documents, along with potentially interesting or useful features of the words such as location information.
Consider a computer program library containing source code. The maintainers of the program library want to design an indexing and retrieval system so that programmers can search for and retrieve useful programs without having to know the programs' names.
You are engaged to design such a system. You may assume that the source code is all written using the same programming language (e.g., C, or Java, or Pascal). Describe the different ways in which you, as an experienced programmer, might want to search for programs, and the program constructs which support those methods of searching. For each query type, are there any mismatches between the granularity of the query (the description of what you are looking for) and the program code (which is all that is available to construct the index), and how can they be bridged? Would you use a boolean, statistical, or probabilistic model of information retrieval, and why? Finally, what recommendations would you make for programmers to make their programs easier for others to search for (i.e., to make the programs easier to index at the level of detail required by the different ways in which you would want to search for them).
End of paper