Presented By
P.K.Surya Prakash
J.Manikantha
P.Raveendra

---

[attachment=11820]
Motivation
• Well we took to this project basically because music rules the whole world in our gen.
• We find loads of people listening to music on the net and it’s no child’s work to store that large database and operate them 24*7.
• Mp3 is a format alternate to the then dominant Wav format which eats up large space.
• So, engineers came up with a simplified version “ Mp3”.
• An Mp3 is a encoded format of a Wav file.
Introduction
• MP3 has changed the way people listen to music on the Internet.
• It was not so long ago that a song in Wav format took hours to download on a 28.8 kbps modem connection and ate up around 50 megabytes of disc space.
• With the same song converted into an MP3 file, download time gets reduced dramatically to around one-tenth the original time and size almost a dozen times lower; while sounding just as good as before.
How does MP3 work?
• As a form of compression, MP3 is based on a psycho-acoustic model ,which recognizes that the human ear cannot hear all the audio frequencies in a recording.
• The human hearing range is between 20 Hz to 20 kHz and it is most sensitive between 2 to 4 kHz.
• When sound is compressed into an MP3 format, an attempt is made to get rid of the frequencies that cannot be heard. As such, this is known as 'destructive' compression.
• The quality of an MP3 file depends on what is called its "sample rate" or "bit rate"
MP3 Bit Rate
• An MP3's bit rate (or sample rate) refers to the amount of audio information (measured in Kb Kilobits) it reproduces per second.
• The higher the bit rate the better the quality, but also the size of the MP3 file is increased, meaning the higher the quality, the less you can fit on your storage device.
Overview of audio compression formats
Inside An MP3 File
• Organization of bit stream in MPEG file
• An MPEG audio file is built up from smaller parts called frames. Generally, frames are independent items. Each frame has its own header and audio information.
• When we want to read info about an MPEG file, it is usually enough to find the first frame, read its header and assume the other frames to be same.
• We may have variable bit rate MPEG files which use so called bit rate switching, i.e. bit rate changes according to the content of each frame.
• The frame header is constituted by four bytes (32bits) in a frame. The first eleven bits of a frame header are always set and they are called "frame sync".
• Hence, one can search through the file for the first occurrence of frame sync.
• Then we read the whole header and check if the values are correct.
• Here is a presentation of the header content. Characters from A to M are used to indicate different fields.
• “AAAAAAAA AAABBCCD EEEEFFGH IIJJKLMM”
• Now we take a look into the contents of the header through a table.
MP3 ENCODER
MP3 Decoder
• Decoding of MP3 audio is carefully defined in the ISO standard.
• Each frame is made up of 1152 samples and there is always a header attached to each of the frames associated in the MP3 file.
• Content in the header and side information for a particular frame is necessary so that decoding is done correctly.
• The first and foremost thing in the decoding procedure is the synchronization of the decoder to the incoming bit-stream.
• Synchronization is the process of finding the position of the first header and the subsequent ones.
• Once this is done, the organization of the encoded data is completely known and the decoding procedure can be performed smoothly
Block Diagram
Retrieving file information
• Retriving file information/Frame unpacking constitutes finding the bitstream header, decoding side information, decoding scale factors.
• Frame size is the number of samples contained in a frame.
• It is constant and is always 384 samples for Layer Iand 1152 samples for Layer II and Layer III.
• For Layer II & III files, frame length is calculated by the formula
FLB = 144 ∗ Bitrate/Samplerate +Padding
Decoding Huffman data:
• First, the frequency lines of the three regions in the big values are decoded and then the small values.
• Decoding is done by using the tables specified in the standard and details on which table to choose is given by the value ‘table_select’ in the side information.
• Once the big values are decoded, the remaining Huffman coded bits are decoded by the value ‘count1table_select’.
• Decoding is done until all Huffman code bits have been decoded or until quantized values representing 576 frequency lines have been decoded, whichever comes first.
Requantization
• Requantization is done separately for both short and long blocks by using a power law and is given in formulas given below.
• Scaling which follows requantization, is done by multiplying the values by the corresponding scalefactors and are stored as scaled frequencies.
Reordering spectrum
• Reordering of spectrum is dependent on the block type used prior to the IMDCT operation.
• This ordering of frequency lines for short windows is done to increase the Huffman coding efficiency.
• If short window is used (block_type=2) then the requantization block would produce frequency lines ordered first by subband, then by window and at last by frequency.
• If long windows are used, then the frequency lines are ordered first by subband and then by frequency
Alias reduction
• Aliasing reduction is done for long block types (ie, block_type!=2).
• The anti-alias block reduces the aliasing that is introduced by the use of ideal non-band pass filter.
• Aliasing reduction is done by merging the frequency lines using eight butterfly calculations for each subband.
• IMDCT
Frequency inversion
The output of overlap add consists of 18 time samples for each of 32 polyphase subbands. Before processing the time samples into synthesis polyphase filter bank , every odd time sample of every odd subband should be multiplied by -1 to compensate for frequency inversion.
Synthesis via polyphase filter bank
The final step in the decoding process is to synthesize the 18 time samples for each of the 32 subbands in each granule, into 18 blocks of 32 PCM samples.