Presentation is loading. Please wait.

Presentation is loading. Please wait.

LOGO Song Identification System Team members: Nguyen Ngoc Tan Ho Vinh Thinh Nguyen Huu Duy Nguyen Hoang Diep Nguyen Trong Dai Le Thanh Tung Supervisor:

Similar presentations


Presentation on theme: "LOGO Song Identification System Team members: Nguyen Ngoc Tan Ho Vinh Thinh Nguyen Huu Duy Nguyen Hoang Diep Nguyen Trong Dai Le Thanh Tung Supervisor:"— Presentation transcript:

1 LOGO Song Identification System Team members: Nguyen Ngoc Tan Ho Vinh Thinh Nguyen Huu Duy Nguyen Hoang Diep Nguyen Trong Dai Le Thanh Tung Supervisor: PhD. Phan Duy Hung

2 Agenda Introduction Project management plan Software Requirement Specifications Software Architecture Design Testing Summary Demo Q&A

3 Introduction  Project information:  Project name: Song Identification System.  Project code: Songfeeler.  Project type: website and mobile application.

4 LOGO Idea

5 Roles & Responsibilities  Supervisor: PhD. Phan Duy Hung  Project Manager: TanNN  Developer team: Mobile Application: TanNN - Teamleader Detector: ThinhHV – Teamleader Website & Database: DuyNH – Teamleader  Test team: DaiNT, TungLT  Document & QA: DiepNH

6 Project management plan  Using waterfall model.

7 Schedule

8 Objective MetricsCommittedActual Start Date8-Sept-2014 End Date19-Dec-2014 Duration93 days Team member66

9 Risks  Missed deadline.  Technologies traps.  Budget.  Human resources.

10 Project Tools & Technologies

11 System Requirements

12 Non-functional Requirements  Usability  Reliability  Availability  Performance

13 LOGO Algorithm

14 Frequency spectrum

15 Short Time Fourier Transform (STFT) Choose a window with finite length Place the window on top of the signal at t=0 Truncate the signal using this window Compute the Fourier Transform of the truncated signal, save results Incrementally slide the window to the right Go to step 3, until window reaches the end of the signal

16 Turn a signal in time domain into frequency domain If we call L s is sample length, F s is sample rate, then applying FFT to the samples will result in L s FFT columns with values correspond to frequencies magnitude: [0, F s /L s, 2F s /L s, …, (L s -1)F s /L s ] So, we must choose L s to balance between time resolution and frequency resolution. Fast Fourier Transform (FFT)

17 We get the frequency spectrum in an 2D array Frequency (L s -1)F s /L s frequency magnitude … …………………… 4F s /L s frequency magnitude … 3F s /L s frequency magnitude … 2F s /L s frequency magnitude … F s /L s frequency magnitude … 0 … Block 0Block 1Block 2Block 3 Block 4… Time Fast Fourier Transform (FFT)

18 Fingerprint Raw sample Original song Fingerprint Compare

19 Required attributes of fingerprint  Temporal-locality  Translation-variant  Robustness  Entropy

20 Spectrogram peaks

21 Frequency spectrum transforms into a set of peaks

22 Simple matching method Each peak in the record and song will be defined by 2 numbers : frequency index and time offset. Use frequency index as the key of matching as following: Record peaks Frequency indexTime offset 8252 18966 36085 …… Song peaks Frequency indexTime offset 82113 189245 82415 360313 …… Match the record with the song Frequency indexRecord time offsetSong time offset 8252113 8252415 18966245 36085313 ………

23 Offset matching chart Each red point has coordinate (SongTimeOffset : RecordTimeOffset)

24 Histogram of TimeDiff = SongTimeOffset - RecordTimeOffset

25 Fast Combinatorial Hashing

26

27

28 How [hash:offset] is calculated Now we use hash instead of only frequency index as the key for matching

29 Rule for pairing peaks Each peak is paired with other peaks inside its Target Zone Fan-out factor is defined to limit the number of pairs

30 What about survived percentage of new structure? p*[1-(1-p) F ] ≈p

31 LOGO Optimization

32 FFT window

33

34

35 Choosing a FFT window function Reference: http://en.wikipedia.org/wiki/Window_functionhttp://en.wikipedia.org/wiki/Window_function

36 Block shifting Shift calculation Try shifting all FFT blocks by a number of samples

37 Software Architecture Design

38 Mobile application  Detect song  History  View song info  View lyrics

39 Detecting process

40 Detector components This is our core component, the brain of our system.  Analyzing component  Searching component

41 Analyzing process

42 The web server has 2 components using 2 different technologies:  ASP.NET MVC  ASP.NET API Web server

43 ASP.NET MVC The first component using ASP.NET MVC which manage songs database which including:

44 Listing

45 Create

46 Edit

47 Delete

48 Attach/Detach

49 ASP.NET API -Responsible for returning song detail information during detecting process -Update detected count for detected song The API component receive song id from detector, and return song information via http request and use json to store data.

50 Database  Main system database includes only one table, this table stores all songs data available in the system used for web server and the detector.

51 Mobile database  Mobile application database includes one table, this table stores data of detected songs in the mobile application and use for showing history function.

52 LOGO TESTING Test Process Test Strategy Test Stage Tools and Environments Testing Execution Test Report

53 V-model Test Model

54 Regssion Testing Unit Testing Integration Testing System Testing Acceptance Testing Developer Tester End-User (our friends) Developer & Tester Test Stage

55 Test process

56 Purpose Tools/Environme nt VersionSource Access to web for user Chrome33Google Firefox 28Mozilla Application testAndroid phone Android OS 4 and above Google Test Report Word 2010Microsoft Test Plan Word 2010Microsoft Test Case Excel 2010Microsoft Test LogExcel2010Microsoft Tools and Environments

57 Defect logs Testing Execution

58 Actual Expect Testing Execution

59 Actual Expect Testing Execution

60 Test Case Stage 1Stage 2 Final PassFailPassFail Android App235280 Website7010800 GUI Test Case462480 Total all Test Case139171560 Test Report

61 LOGO DEMO


Download ppt "LOGO Song Identification System Team members: Nguyen Ngoc Tan Ho Vinh Thinh Nguyen Huu Duy Nguyen Hoang Diep Nguyen Trong Dai Le Thanh Tung Supervisor:"

Similar presentations


Ads by Google