Audio Coding MPEG 1 Layers I III MPEG

Audio Coding MPEG 1 Layers I, III MPEG 2 MPEG 4 Sherida Subrati Anthony Caliendo

Overview • Explanation of Codecs ØMPEG 1 – Layer I, III (Differences) ØMPEG 2 – Basic Overview ØMPEG 4 – Possible Applications • Applications & Sound Samples Used • Results & Explanation ØFile Size, Bitrate, & Quality ØWaveform Comparison • Summary & Questions

Sub-Band Coding Overview • Size of sub- bands varies • Varying application of psychoacoustic model

MPEG 1 – Layer I & II • Time Frequency Mapping Ø Polyphase Filter Bank Ø 32 Equal Bands • Psychoacoustic Model Ø 512 -point FFT & 1024 -point FFT respectively Ø Tonal & Noise Masking • Quantizer Ø Scale Factor: 6 bits Ø Layer II – Allows 3 successive scale factors & uses 1 -3 depending on how much they differ

MPEG 1 – Layer I & II Diagram Images from Peter Noll MPEG Digital Audio Coding Standards

MPEG 1 – Layer III • Time Frequency Mapping Ø Switched Hybrid Filter Bank Ø 32 sub-bands further sub-divided using a 6 or 18 -point DCT • Psychoacoustic Model Ø Variable FFT Ø Tonal & Noise Masking • Quantizer Ø Non-uniform Scale Factors Ø Huffman Coding, Bit Reservoir, & Iterative Analysis

MPEG 1 – Layer III Diagram Images from Peter Noll MPEG Digital Audio Coding Standards

MPEG 2 – General Overview • 5. 1 Channel Support • Advanced Audio Coding (AAC) Ø Optional Preprocessing Ø Bit-stream Formatter Ø Prediction – helps to optimize quantizer Ø Noiseless Coding Ø 3 Profiles • Main – Variable length DCT, noiseless coding, etc. • Low Complexity – No temporal noise shaping & time domain prediction • Sampling Rate Scalability – preprocessor allows for sampling rates of 6, 12, 18, & 24 KHz

MPEG 4 - General Overview • Consists of all previous MPEG iterations • Uses 3 Core Coders ØParametric coding for low bit rate speech ØAnalysis-by-synthesis for medium bit rates ØSub-band/Transform coding for high bit rates • Low Delay (LD) Encoding / Decoding • Quality Scalability

MPEG 4 - Diagram

Applications & Sound Samples • Applications Ø AVI 2 MP. EXE Ø LAMEwin 32 Ø Nero MPEG 4 AAC Ø Goldwave • Hardware • Sound Samples Ø PCM 16 -bit Stereo 44. 1 KHz • Clubbed to Death (Kurayamino Mix) – Rob D • Man Who Sold The World Nirvana Ø PCM 8 -bit Mono 44. 1 Khz • Voice Sample Ø Pentium III – 1. 0 GHz Ø 512 MB RAM Ø Win 2 K SP 3

Results – File size VS Bitrate

Results – Encode Time VS Bitrate

Results – Quality VS Bitrate

Sample Sounds • Music Sample Ø Original Sound Ø Sample 2 Play list Ø S 2 -M 4 LT-064 S Ø S 2 -M 4 LT-080 S Ø S 2 -M 4 LT-096 S • Voice Sample Ø Original Sound Ø Sample 3 Play list Ø S 3 -M 4 LT-016 M Ø S 3 -M 4 LT-024 M Ø S 3 -M 4 LT-032 M

Sample Waveforms – S 2 -64

Sample Waveforms – S 2 -128

Sample Waveforms – S 3 -64

Sample Waveforms – S 3 -96

Summary • MPEG 1 – Layers I, II have limited options & are not size versus quality efficient • MPEG 1 – Layer III offers excellent quality at low rates but has large overhead • MPEG 2 – Much more comprehensive • MPEG 4 – Encompasses all previous iterations & has new capabilities to increase its lifespan