Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                
Scribd Logo
Upload

Welcome to Scribd!

  • 14 views

    What Is Shazam?

    Uploaded by

    Popa Mihai
    Shazam is a mobile application that listens to music and compares the spectrogram (fingerprint) of the audio against a large database of songs. It takes a 10 second sample, identifies salient peaks in the spectrogram, and pairs these peaks to form landmarks. These landmarks are used as hash keys to efficiently search the database and find matching songs based on identical landmarks occurring at similar time offsets from the beginning of each song. If a match is found, Shazam provides information about the song and offers options to purchase or stream it.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd

    What Is Shazam?

    Uploaded by

    Popa Mihai
    14 views22 pages
    Shazam is a mobile application that listens to music and compares the spectrogram (fingerprint) of the audio against a large database of songs. It takes a 10 second sample, identifies salient peaks in the spectrogram, and pairs these peaks to form landmarks. These landmarks are used as hash keys to efficiently search the database and find matching songs based on identical landmarks occurring at similar time offsets from the beginning of each song. If a match is found, Shazam provides information about the song and offers options to purchase or stream it.

    Original Description:

    info about shazam

    Original Title

    9_Shazam

    Copyright

    © © All Rights Reserved

    Available Formats

    PDF, TXT or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    Shazam is a mobile application that listens to music and compares the spectrogram (fingerprint) of the audio against a large database of songs. It takes a 10 second sample, identifies salient peaks in the spectrogram, and pairs these peaks to form landmarks. These landmarks are used as hash keys to efficiently search the database and find matching songs based on identical landmarks occurring at similar time offsets from the beginning of each song. If a match is found, Shazam provides information about the song and offers options to purchase or stream it.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    14 views22 pages

    What Is Shazam?

    Uploaded by

    Popa Mihai
    Shazam is a mobile application that listens to music and compares the spectrogram (fingerprint) of the audio against a large database of songs. It takes a 10 second sample, identifies salient peaks in the spectrogram, and pairs these peaks to form landmarks. These landmarks are used as hash keys to efficiently search the database and find matching songs based on identical landmarks occurring at similar time offsets from the beginning of each song. If a match is found, Shazam provides information about the song and offers options to purchase or stream it.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    of 22

    What

     is  Shazam?  

     Mobile  phone  based  applica5on  that  listens  


    to  music  and  compares  against  a  database.    
     
    Shazam’s Method

    ! Ideas
    ! Take advantage of music local structures
    ! Find salient peaks on spectrogram
    ! Pair peaks to form landmarks for comparison
    ! Efficient search by hash tables
    ! Use positions of landmarks as hash keys
    ! Use song ID and offset time as hash values
    ! Use time constraints to find matched landmarks
    Technology  

    ! Takes  10  second  sample  


    ! Catalogs  the  spectrogram  (fingerprint)  
    ! Compares  against  large  database  
    ! If  match  is  found,  
    ! Gives  song  informa5on  
    ! Offers  immediate    purchase  
    Shazam-Industry leader in audio fingerprinting

    ! For each audio file, generate reproducible landmarks


    ! Each landmark occurs at a time offset
    ! For each landmark, generate a “fingerprint” tag that
    characterizes its location
    Spectrogram and fingerprint

    ! Spectrogram axes
    ! x axis: time
    ! y axis: frequency
    ! z axis: intensity
    ! identify frequency of peak intensity

    5
    Spectrogram and fingerprint

    ! Spectrogram axes
    ! x axis: time
    ! y axis: frequency
    ! z axis: intensity
    ! identify frequency of peak intensity

    6
    Shazam-Industry leader in audio fingerprinting

    ! Do same for sample

    ! slide the mask to match a significant number of points

    ! Generate list of matching fingerprints

    ! timedb–timesample= Constant
    Shazam: Landmarks as Features

    Pair peaks in
    target zone to
    Spectrogram form landmarks

    •  Landmark: [t1, f1, t2,


    Salient peaks of f2]
    spectrogram •  24-bit hash key:
    •  f1: 9 bits
    •  Δf = f2-f1: 8
    bits
    •  Δt = t2-t1: 7
    bits
    •  Hash value:
    (Avery Wang, 2003) •  Song ID
    •  Landmark’s
    start time t1
    Shazam: Landmarks as Features

    Pair peaks in
    target zone to
    Spectrogram form landmarks

    Salient peaks of
    spectrogram

    (Avery Wang, 2003)


    Shazam: Landmarks as Features

    Pair peaks in
    target zone to
    Spectrogram form landmarks

    •  Landmark: [t1, f1, t2, f2]


    Salient peaks of •  24-bit hash key:
    spectrogram •  f1: 9 bits
    •  Δf = f2-f1: 8 bits
    •  Δt = t2-t1: 7 bits
    •  Hash value:
    •  Song ID
    •  Landmark’s
    start time t1

    (Avery Wang, 2003)


    hash

    ! anchor point and target zone


    ! for each point in target zone
    ! hash: f1, f2, Δt
    ! robust to time shifts, noise

    11
    search

    ! search for all matching songs

    12
    2d plot of frequency hits

    ! if a specific song is hit multiple times:


    ! 2d plot of frequency hits
    ! x axis: time from the beginning of the track, whose
    frequencies appears in the song
    ! y axis: time from the beginning of the track, whose
    frequencies appears in the sample
    ! if the files match, matching features should occur at similar
    offsets from the beginning of the file

    13
    Shazam-match

    ! if the files match, matching features should occur at similar


    offsets from the beginning of the file
    Shazam-match algorithm

    ! matching features should occur at similar time offsets


    ! compute offset: dk = yk – xk
    ! histogram of dk
    ! sort dk
    ! scan for a cluster of values
    ! match: presence of a statistically significant cluster

    15
    Shazam-no match
    Hash

    ! no special assumption about the format of the hashes


    ! hash
    ! to avoid having too many spurious matches
    ! being reproducible
    ! scanning:
    ! matching hashes to be temporally aligned
    ! score
    ! number of matching and time aligned tokens

    ! from distribution of scores of false positive


    ! estimate threshold score

    17
    Technology

    ! Pair of peaks: peak + anchor


    ! Hash table searching O(l), collision reduced by having extra key
    by orders of magnitude
    ! Robustness to interference?
    Societal Impact

    ! Introducing people to new music.


    ! Hear a song once, have it forever
    Economic Impact

    ! Improves music sales


    ! Music sales are a convenient impulse buy
    ! More efficient advertisement of music.
    ! http://www.youtube.com/watch?v=ppJAkN4m9bY
    Conclusion  

    ! Turns  the  world  into  a  record  store  


     
    Riferimenti

    ! A.Wang, “An industrial-strength audio search algorithm”, ISMIR


    2003

    22

    You might also like