Tutorial Questions (Mpeg Audio Compression, Mp3)
CSE 3004Y Multimedia Systems Tutorial Questions (MPEG Audio Compression, MP3)
Question 1 Consider the MPEG-1 Audio standard. a) Determine how many complete stereo songs with a 22 kHz bandwidth and 16 bit/sample with duration of 3 minutes is possible to store in a disk with 200 Mbytes if the coding is performed with MPEG-1 Audio Layer 3 to reach CD transparent quality. b) Why does this audio standard use the DCT with overlapping window? c) Which main factors would you take into account (at least 3) to select one of the MPEG-1 Audio Layers to code the audio for a certain application? Question 2 a) From a high-level viewpoint, MP3 works as follows: 1. Divide the raw audio samples into timeslices (group of samples). 2. Apply a Modified Discrete Cosine Transform to each timeslice. 3. Use a psychoacoustic model to predict the masking thresholds for each frequency coefficient that was output by the MDCT. 4. Don’t code the frequencies that won’t be heard. 5. For the frequencies that are audible, re-scale the frequency coefficients so as to quantize them more coarsely. 6. Huffman code the quantized coefficients. Describe the purpose of each step?
b) When MP3 is stored in a file, many good-quality MP3 encoders give the option of generating a constant bitrate (CBR) MP3 or a variable bitrate (VBR) MP3 file. For CBR and VBR files of the same size, the quality of the VBR file will be better. Explain why this is so. c) If a 128Kb/s MP3 file is decoded to an uncompressed WAV file, then later recoded again as a 128Kb/s MP3 file, the quality will degrade slightly. Explain why this is so. Question 3 (a) Briefly describe the psychoacoustic compression approaches that are exploited in MP3 audio compression.
(b) Explain how Sample Rate and Sample Size affect audio quality when converting analogue sound to digital form. Include in your explanation references to quantization error and Nyquist’s theorem, use diagrams to support your answer. Question 4 (a) Explain the following
Question 1 Consider the MPEG-1 Audio standard. a) Determine how many complete stereo songs with a 22 kHz bandwidth and 16 bit/sample with duration of 3 minutes is possible to store in a disk with 200 Mbytes if the coding is performed with MPEG-1 Audio Layer 3 to reach CD transparent quality. b) Why does this audio standard use the DCT with overlapping window? c) Which main factors would you take into account (at least 3) to select one of the MPEG-1 Audio Layers to code the audio for a certain application? Question 2 a) From a high-level viewpoint, MP3 works as follows: 1. Divide the raw audio samples into timeslices (group of samples). 2. Apply a Modified Discrete Cosine Transform to each timeslice. 3. Use a psychoacoustic model to predict the masking thresholds for each frequency coefficient that was output by the MDCT. 4. Don’t code the frequencies that won’t be heard. 5. For the frequencies that are audible, re-scale the frequency coefficients so as to quantize them more coarsely. 6. Huffman code the quantized coefficients. Describe the purpose of each step?
b) When MP3 is stored in a file, many good-quality MP3 encoders give the option of generating a constant bitrate (CBR) MP3 or a variable bitrate (VBR) MP3 file. For CBR and VBR files of the same size, the quality of the VBR file will be better. Explain why this is so. c) If a 128Kb/s MP3 file is decoded to an uncompressed WAV file, then later recoded again as a 128Kb/s MP3 file, the quality will degrade slightly. Explain why this is so. Question 3 (a) Briefly describe the psychoacoustic compression approaches that are exploited in MP3 audio compression.
(b) Explain how Sample Rate and Sample Size affect audio quality when converting analogue sound to digital form. Include in your explanation references to quantization error and Nyquist’s theorem, use diagrams to support your answer. Question 4 (a) Explain the following