Organisational questions:
- What should our Camera/desk setup look like?
- A: An external camera, e.g., a smartphone camera, at your side, so that you're visible and what is on your screen. You must log in twice.
- When do we get the link to join the exam?
- When do we have to upload the student ID card in Opal?
- A: we just check the ID live during exam check-in, no upload required
- What if we do not have ID, since we are giving exam from home country?
- Can we show soft copy of immatriculation and passport to you?
- Do we have to join to big blue button before exam?
- A: Yes, please be in your specific exam webroom which belongs to the exam group you will be enrolled, 1 hour before the exam.
- At what time?
- A: 3 pm / 15:00 to authenticate, the exam starts at 4 pm / 16:00
- Would it be possible to draw ?
- A: on a sheet of blank paper yes, but there will be no digital drawing, on the TUCexam website
- Will the exam be of 90min?
- Will the exam be fully done in tucexam/opal or are there parts which we need to write on paper and scan + hand in later?
- How will we be able to insert special signs, logarithms (e.g., sigma, floor)?
- A: in all cases where special signs are needed, we provide examples on how to write it.
- Would we have log questions?
- A: yes, minor equations containing log may be requested from Parts "Information Theory" or "Compression Techniques"
- Is the exam completely closed book or is there any option to have an equations sheet or related?
- A: closed-book, no equation sheets
- Are we allowed to use clear paper in the exam to calculate or draw trees?
- A: Yes, but in the end you must type in all answers to the text inputs on TUCexam.
- Use of a calculator is allowed?
- If we must be present an hour earlier for the exam, are we allowed to move around the room or go to the toilet between 3-4pm or must we stay at the desk?
- Can we use a list with the formulas during the exam?
- Would be 60 min enough for the exam?
- A: Time will be quite tight but some students finish earlier
- Would you please explain again about the external camera and the time briefly?
- see scheme and previous questions
- Do we have to show our room / desk?
- During check-in yes, we request you to pan the camera around the room to verify that nobody else is there.
- Can we use two screens during the exam? I mean the monitor as another secondary screen?
- Yes, but please make sure all monitors are visible in the field of view of the camera.
- Will the other students be able to see our room etc.?
- No, only we as supervisors are able to see all students webcams at once, students don't see any webcams at all.
- How can we draw tree diagrams, e.g., Shannon-Fano or Huffman?
- We have prepared some edit fields to fill in, but the lines between them are omitted. You can see the pattern/structure that you would normally have to draw though.
- Just to be clear, any kind of calculator is allowed?
- No, but the calculations will be simple and no float numbers.
- What happens if we have internet problems during the exam?
- Contact your exam proctor in your BigBlueButton webroom via private chat.
- Is my answer saved when problem in internet appears?
- Yes, but you might not see all your old answers when you log in again. Skip them and continue.
- Can we go back to the older qustions to review them?
- Yes, you can move freely forward and backwards between questions.
- Q: Do we get any minus point for wrong answer?
- In general no. But if it will be the case, we will inform you about it in the task description.
Subject related questions
A: Information Theory
- Q1: Tutorial 2. Take 8 question a. How to quantize a mid-rise signal?
- A: mid-rise is rather a quantizer not a signal, x-axis, q is on y axis
- Q2: How we are calculating memory space fixed length ? are we taking same probability for each symbol? So the formula is log k (base 2) always and in variable length memory space how the calculation is differing?
- A: In a fixed length code you the memory needed is just the product of code length [Bit] * number of codewords
- In a variable length code it's not as easy to calculate. You would need to sum up the number of Bits for all codewords individually.
- Q3: What is the difference between a quantization error and a reconstruction error?
- A: Reconstruction error occurs due to errors made during quantization. In other words, quantization errors show up in form of reconstruction error after reconstructing the signal.
- The cause is the quantisation, but it will show itself in reconstruction
- Q4: For Tutorial 2, Task 8a): What will be the delta, and how we get these numbers 3 -3 ....?
- A: You don't need formulas, you check it graphically.
- Q5: What is message with max and the minimum entropy for {0,1,2,3,4,5,6,7,8,9}? How do I know how many characters should be used in the message?
- A: number of characters will be given in task
- here: min = max because characters are equally often
- Q6: What are the differences in the coordinate systems between a digital and analogue signal?
- Analoge: y-axis is amplitude in Volts, x-axis is time in seconds
- digital domain: x-axis (time, but could also be number of samples), y-axis: quantisation levels q (unitless)
- Q7:in lecture 4, page 3 in table ,why the ICs are with minus-?
- A: yes, because log(1/pi) = -log(pi)
- Mistake in the script, page 44, in the table --> Information content should be positive values
- Q8: dynamic range, headroom and footroom?
- has been skipped in the exercises
- also not relevant in image encoding
- Q9: How can quantisation error be minimised?
- increase the quantization bit depth
- Q10: Haming distance between 00110 Λ 01010 Λ 01100.....change between digits is 3? so hamming distance should be 3?
- Oral explanations were given.
- Q11: Can you explain once the calculation?
- Q12: but in Q10 .is 2,but why is 3?
00110
01010
-------
01100 = 2
- Q13: I did not understand why min and max is equal ,because min is when all chars are equal for ex:0000000... but here they are all different and make max not min.
- A: min/max entropy follow-up question
- in theory it depends on you alphabet, if zeros occur more --> often low entropy and information content, e.g. "e", "n" occur frequently in english --> low IC, words with lots "e"s have low entropy
- Q14: But those zeros are different between each other?
- A: No, they are the same type of zeros. The set notation {...} is just another way of providing the statistics of probability of occurence of the different characters.
- Q15: Basically the entropy on these examples depends on the message given according to the probability of occurrence of every character?
- Q16: When the task says: "Write the message with max and min entropy for {00011123}", should we calculate Hmax and Hmin or just write a message?
- A: You just write the message, you aren't required to calculate anything but you need to know how to find the messages.
B: Compression Techniques
- Q1:When coding with Huffmann, how do we choose which signs from the coding table to use, if there are more than 2 with the same number?
- A: You always go from bottom to top. 1. Sort your given alphabet in a descending order based on the absolute number of occurances. 2. You go from the bottom of the table to the top and take pairs of two. 3. You sum the pair up into a node and treat them as one from now on. You can imagine the node like a new entry in your table, i.e., compare the node with the number of occurances of the remaining characters. Add the node to your tree when its value is among the lowest number of occurances like you would with any other character.
- Video: https://www.youtube.com/watch?v=JsTptu56GM8
- Q2: Does it matter which sides the parts of the tree are?
- Q3: Difference between static and dynamic pattern substitution? and dynamic pattern substitution advantages over static pattern substitution? (pp. 36-37)
- A: Static pattern substitution: The rules that you set in the beginning will always be applied in every compression you do. It's not adaptive to the messages you compress. All text you ever compress will follow the rules. Dynamic pattern substitution: The rules are set up and adapt based on the messages you compress.
- Advantages: Static patterns are good for small alphabets, dynamic - good for longer texts with many unpredictable variations (?)
C: Image Encoding
- Q1: CMYK is used for printing media and RGB is used for Light emitting media. How can we justify these statements?
- A: CMYK is a subtractive colour space and requires a white background that is illuminated with a natural or light and hitting the pigment, creating the impression of the correct colours. RGB is an additive colour space and in this scenario, you start with a black background of the monitor to which components of the colours are added by adding light on the screen and create the impression of the image.
- This is a cool video explaining this topic, just switch on the english subtitles: https://www.youtube.com/watch?v=r0jXfwPQW9k
- Q2: Quantization matrix for JPEG. About low frequencies strong or less than high frequencies comparison
- A: question quite vaguely put. Oral explanations given using lecture slides.
- Q3: In Tutorial 6, Task 1c). Do we need to calculate the inverted matrix or will it be given in the exam?
- A: No, you don't have to calculate inverted matrices.
- How do we calculate all colours available based on Bits?
- Q: (2^Bits)^number of channels
- If is asked for quatization tables to show high compression or other properties, we should use any number in tables just show this properties or special number? mean for top left low number for bottom high nubmer?
- No hard boundaries but roughly: 1 is no compression, 5-30 mild compression, 50+ strong compression.
D: Video Encoding
- Q1: "How are I-frames coded? Name steps of the coding process." Is the answer JPEG compression...?
- Q2: Will you write in the question the unit you want the answer in? If we do not turn them, will we get less points?
- Q3: Can u explain Two-dimensional logarithmic search working ?
- A: Oral explanations were given using lecture slides.
