Q1.a)Briefly explain the advantage of logarithm b)the main steps performed in a audio coder.

quantification.

Ans: logarithm quantification provides quantification errors for low amplitude values. b)The advantage of backward adaptive

quantification .

Ans:Don’t require transmission and has lower delay .

Q2.a)The difference between Lossless and lossy compression . Ans:Lossless compression reconstructs original signal

while lossy compression produces an accurate representation of the original signal . Q3.a)The source filter model of speech production . Ans:The source filter model consists of filter,the input of

which is either a periodic pulse train or random noise .The source is chosen using a switch .This model is used synthesis speech and model speech production . b)With regard to human hearing ,define simultaneous masking . Ans:This is where a tone is masked by another tone or set of tones occurring in the same frequency region as the single tone and at same time. c)Which type of speech signal are most suited to

linear prediction and explain why.

Ans:Voiced speech signals are most suited due to the presence of short term correlations. d)How is the performance of speech and audio coders evaluated?Why SNR\subjective measures? Ans:The performance can be evaluated using subjective and objective measures.This is because there is no fixed correlation between objective and subjective error ----a large mathematical error such as measured with the SNR may not result in a perceivable error. Q4.a)The method represented the LP excitation

signal in 3 broad methods . Ans:Direct ---G721 uses ADPCM ; Analysis-by-Synthesis---Quantise each sample ; Parametric ---Modelling the excitation using e.g. pulses for voiced speech and random noise. b)What is the pitch period of a speech signal ? Ans:the pitch period of a speech signal corresponds to

the measured period of the quasi-periodic voiced sections of a speech signal.This periodic nature is caused by the periodic opening and closing of the vocal folds in the vocal tract. Q5.a)MP3 is based on the perceptual coding paradigm.Explain how audio coders operating. ANS:Audio coders operate by using a psychoacoustic model.The most important components are represented more accurately than the masked components. bands. Perceptual Model:Calculate a psychoacoustic model for the input spectrum. Quantization and coding:Quantise the spectral

components falls below the masking threshold . Encoding of Bitstream:Use entropy ending to further reduce the bit rate. c)two key difference between MP3 and AAC? ANS:TNS ; Joint stereo coding . Q1B:a)the difference between a grey scale digital image and a colour digital image . ANS:Grey scale image is black and white .Only one colour component per pixel with a range of values form 0

to 255.colour images have multiple components per pixel ,representing red,green and blue. b)the purpose of transforming in image compression ? ANS:transforming aims to compact the energy . c)What criteria should be used when choosing the resolution for a digital image ? ANS:the subjective distortion is minimised. d)compare RGB colour space with YCrCb colour space. ANS:Y is made up of all R,G,B components.Cr and Cb is

one of the colour components.They represent colour. e)explain the term critical flicker . ANS:the minimum frame rate below which the human eye will detect flicker. Q2B:a)Which coefficients are the most important in DCT image compression ? ANS:The coefficients with the highest magnitude .These

are the first few coefficients obtained from a 2D DCT of the image . b)What is the zigzag scanning pattern ? ANS:This is designed to order the coefficients in magnitude.It will lead to long sequences of zeros. Q3B:a)Compare the DWT-based image compression and DCT-based image compression . ANS:DWT-based image compression will result in blurred distortion whereas DCT-based image compression will result in blocky distortions.

b)two advantages of JPEG20xx?

ANS:Lower bit rates for a given quality ;increased scalability .

c)describe ROI and why it is used in JPEG20xx.

ANS:It is used in JPEG20xx to allow fast reconstruction of important parts of an image .

d)The two components that need to be compressed in video comprehension .

ANS:The image frame and the temporal info. I.e. Motion .

e)Describe two problems with differential quantisation.

ANS:Two problems are random access and transmission errors.Avoid by using block based differential encoding.

Q4B:a)Interframe coding:code the info. between the frames using e.g. differential quantization.

Itraframe coding:code the info. within the frame.

Interframe coding leads to Lower bit rates than intraframe coding.

d)Motion estimation :Determines how blocks of pixels have moved from one frame to the next.

Motion compensation:Use motion vectors to predict the current frame.

Motion estimation and compensation is used to code the temporal info. of video and reduce bit rates.

Q5B:a)Explain the 3 step algorithm used in motion estimation and advantage of comparing with EBMA? ANS:1:finding the closest match.

2：center the search block of 1/2 the size as step 1 at vector .Perform 8 matches around this point .

3：repeat step 2 for a search region 1/2 the size of step 2.Choose the vector giving the closest match.

This has the advantage of a much shorter search time. c)The difference between I,P,R frames. ANS:I---Interframe coded frames P---Backward predicted frames

B---Bi--directionally prediction frames

For a given distortion rate ,using more B frames leads to a lower bit rate.

d)ANS:MPEG-4：object based MPEG-1/MPEG-2:frame based

An advantage of MPEG-4 is that is has more scalability ,more interactivity and lower bit rates.

Q6:a)Describe the JPEG encoding using a flow diagram.

ANS:image---------------------8*8 blocks---(-128)---DCT on blocks---Quantize ---zigzag scan---runlength encoding ---Huffman encoding ---compressed image

Q7:What are the drawbacks in JPEG?How do we

improve this scheme ?

ANS:Blockiness ,poor quality in text compression.These drawbacks are mitigated in JPEG20xx using larger block sizes.JPEG20xx is also solves the problem of poor text quality. [Y U V]矩阵：

0.299 0.587 0.114 -0.147 -0.289 0.436 0.615 -0.515 -0.1

高频脉动矩阵： 0 8 2 10 12 4 14 6 3 11 1 9 15 7 13 5 20xx年

Q1:a)the difference between Analogue and Digital signal.

ANS:Analogue---continuous ,Ex:speech signal Digital---Discrete;Ex:JPEG image file

b)explain why LZ77compression is not be appropriate ?

ANS:LZ77 require long files to find redundancy .

c)compress the sequence of character (Y) using LZ77

ANS:input sequence :Y=ABCABCAB