cbtgeek:
I formatted my pc today so I had a little time today when I was setting my codin environment I thought about this contest for a while now and I came up with an ideal on how to tackle the issue
First if you have been searching for a sound recognition library bro am sorry to say this to you "you no go even smell last position for this contest" coz sound recognition only recognize a as a. Not to a particular person
Now let's start the lesson:
I know I am ordinary guy like that with only my waec as my highest degree which I got only three credits
But let forget about that now you are about to learn something that will give you million of naira

Becoz of the contest I wouldn't go on full details I will only touch the explicit parts

I remembered on linux I compile one music player. I don't remember the name of the software. If you play a song with it later if you edit the file name and all the tags it will still recognize the song and give you the name
Now how does this work it is simple wave or sound fingerprinting:
What the hell is sound fingerprinting:
I know this will sound strange but any computer science student here will be familiar with the terms wave fingerprinting. It is all about concentrated digital summary of a sound samples through a sound signal

Now to brake it down ex:
Let take "cum" and "come" to a normal human ear they sound same and also to normal mic but when u sample the two sound very well you notice one will have a higher pitch or shorter sample length
Now if you have a sound fingerprinting of these too words by me even if obj say it like me it will still not match my sample perfectly

Now how does this help in the contest :
If you can get at lest 10 interview of buhari like where he said, nigeria,nigerians,corruption etc
And produce there sample bro you will clock up to 95% and seun go give u ur 50k sharp sharp

I know some pple will criticize me "abi moni no dey sweet u " it is not all about the moni it is all about what you learned from every new challenge but if you are wise I know I dnt explain every thing you can go to google and start research
No body is above mistake If you think u have a better algorithm plz share it I will love to learn thanks

I think audio or sound Fingerprinting will only work when you are matching exactly the same sound / audio file.

nollyj:


I think audio or sound Fingerprinting will only work when you are matching exactly the same sound / audio file.

Yes and no bro voice recognition can only recognize a as a b as b c as c and etc not a particular voice

Now what sound finger printing does when it hash the value of highest frequency making it possible that voice with same pronunciation can have different fingerprinting

Now you can find a long speech of pmb and slice off some of his regular words like nigeria,corruption,youth,boko haram etc and finger print this words if you are given a audio file ,then u match it against ur hashed words . Like every ten seconds of file against ur hashed words if it match it will return 0 else 1

With this method u accuracy is about 97%

Yah! this thread don turn to UNIVERSITY O! SEE ALL THEM BIG BIG PROFESSORS ABOVE ME WITH THEIR BEARDS REACHING GROUND!

cbtgeek:

Yes and no bro voice recognition can only recognize a as a b as b c as c and etc not a particular voice

Now what sound finger printing does when it hash the value of highest frequency making it possible that voice with same pronunciation can have different fingerprinting

Now you can find a long speech of pmb and slice off some of his regular words like nigeria,corruption,youth,boko haram etc and finger print this words if you are given a audio file ,then u match it against ur hashed words . Like every ten seconds of file against ur hashed words if it match it will return 0 else 1

With this method u accuracy is about 97%

It is easier said. Run a simple example of sound fingerprinting with audio file record under different situation but with same voice and show me if the result is 97% match.

Understand that sound fingerprinting is almost the same like fingerprinting. In sound/audio, there are many factors that differentiate two different songs from the same person. Fingerprinting works when there is a matching pattern in frame windows of two different audio files.

At least I have learnt something new - that you can actually take the fingerprint of Sai Baba's voice. . .and make a record label of it

dhtml18:
At least I have learnt something new - that you can actually take the fingerprint of Sai Baba's voice. . .and make a record label of it

Where's it Bro

You mean the money bah? It is safely locked up on nairaland's server and only the oga at the top has access - others can only try. There is no need to weep I assure you, maybe out of pity - Seun go just dash you the money.

I don search web tire maybe I can find anything opensource, but no way. Even my Indian friends are falling my hand, this one don pass me o.

I know that am not perfect but lets wait and see who can win this without finger printing
Hashing,fft,wave length,amplitude,frequency, are all your friend in the contest if some body can create a voice recognition that can detect if a bird is a male or female now tell me if fingerprinting is not the answer

cbtgeek:
I know that am not perfect but lets wait and see who can win this without finger printing
Hashing,fft,wave length,amplitude,frequency, are all your friend in the contest if some body can create a voice recognition that can detect if a bird is a male or female now tell me if fingerprinting is not the answer

bros , write something and stop all these unsolicited lectures... It's getting irritating

No solution yet, but plenty lectures. . . .

jacob05:
bros , write something and stop all these unsolicited lectures... It's getting irritating

write what to prove what do i look like a noob to you if you want the code u can get it for 40k but after the contest you are not allow modify or share it

cbtgeek:


write what to prove what do i look like a noob to you if you want the code u can get it for 40k but after the contest you are not allow modify or share it

40k? Why don't you collect 50k from Oga Seun by submitting ur code to him.. you'd earn 10k more..

See boys trading for codes them never fit write o! egba mi o!

bcoz seun na biz man him go share the code

Yesterday I decided to try my hand on this project. knowing that I am a beginner programmer, I am more interested on what I will learn. I did some research and read some few papers on this topic.

I decided to make a rough trial demo of 3rd Party Libraries I found. My intended language was Java.

I first wrote a desktop application using a Java library called Recognito. The link is here if you want to read more about the library.

I will say that it is a good java library for Text Independent Speaker Recognition in Java. I integrated it to my application and checked the usefulness

The NetBean Java application is attached below. I obtained the following results in four different Buhari speeches



The first two results show the positive matching result of different audio speeches from President Buhari.

The third result was 100 % positive when you match the same audio speeches from Buhari.

Finally, the fourth result is a false negative result. When I dig into to find out why the result return negative, I realize the following factors can prevent detection of a speaker voice.

1. The encoding software used to convert from any file type to .wav file

2. Noise in an audio file

3. Quality of the audio file

It is also important to understand that there is a difference between Speaker Recognition and Speech Recognition. According to wikipedia - “Speaker recognition is the identification of a person from characteristics of voices (voice biometrics). It is also called voice recognition. There is a difference between speaker recognition (recognizing who is speaking) and speech recognition (recognizing what is being said).”

I went on to test Audio /Sound Fingerprinting in android app

(The file size is big to attach here. I will look for a way to share to anyone that wants to download)

Java Sound Fingerprinting Library

After I finished the demo for Desktop, I decided to port a new application to android mobile since I like developing for android.

I decided to try sound / audio fingerprinting java library called Musicg. The link is here if you will like to read more. https://code.google.com/p/musicg/

Th android application has the option to add audio file in three different ways. The display of the app is as shown below.



You can download the source code for the android application.

Sound / audio fingerprinting result from android app

I was able to achieve 100 % positive match using fingerprinting for the same audio file. With other different audio file with same speaker returned no match.

In reality, this is a difficult area with the level of my programming skill. There is also the possible to obtain better result with other programming language that I did not test.

No doubt I have learn something new. Time to go back where I belong (Android Development).

E ma wa o, see all the codes above me o - boys are really angry!
Oya make me sef open my android studio duplicate the code above with

adb -nairaland clone fingerprint.nollyj.nairaland.com

To kickstart my own clone (in case adb fails me) Nollyj abeg put that your code for github sharperly make i use am do ehm template sharp sharp.

Seun, will possible Buhari source file(s) come from radio, television (press conference), phone, or live mic recordings? Or are all sound filters and fidelities possible?

I need a break from 3d printing and web crap

Let me explain once and for all the kind of buhari your app must capture:
- Buhari's voice when he was a baby (crying)
- Buhari when he was a teenager
- Buhari before He became president
- Buhari when he had a cough and sorethroat
- Buhari if he is speaking with his mouth full (but he shouldn't anyway)

And it must be able to detect it was buhari:
- when he speaks in a party with jamz rocking
- when his voice is broadcasted from a megaphone across the street
- when his recorded voice is being played with my nokia 3310
- my daddy's record discs

Under the following moods:
- happy
- sad

Under the following moods:
- crying
- singing

In the following dimensions:
- Mono
- Stereo
- 3D surround

In the following weather conditions:
- Heavy rain with winds blowing
- Thunderstorm situations


Well. . . . .if your code can fulfill all that, I see no reason why you should not win the award

- Happy coding

dhtml18:
Let me explain once and for all the kind of buhari your app must capture:
- Buhari's voice when he was a baby (crying)
- Buhari when he was a teenager
- Buhari before He became president
- Buhari when he had a cough and sorethroat
- Buhari if he is speaking with his mouth full (but he shouldn't anyway)

And it must be able to detect it was buhari:
- when he speaks in a party with jamz rocking
- when his voice is broadcasted from a megaphone across the street
- when his recorded voice is being played with my nokia 3310
- my daddy's record discs

Under the following moods:
- happy
- sad

Under the following moods:
- crying
- singing

In the following dimensions:
- Mono
- Stereo
- 3D surround

In the following weather conditions:
- Heavy rain with winds blowing
- Thunderstorm situations


Well. . . . .if your code can fulfill all that, I see no reason why you should not win the award

- Happy coding

exactly my observations too.

jacob05:
exactly my observations too.

Thank you o jare.

AAinEqGuinea:
Seun, will possible Buhari source file(s) come from radio, television (press conference), phone, or live mic recordings? Or are all sound filters and fidelities possible?

I need a break from 3d printing and web crap

So this question does not even come into it at all. Na why I write my epistle above be this.

dhtml18:
Let me explain once and for all the kind of buhari your app must capture:
- Buhari's voice when he was a baby (crying)
- Buhari when he was a teenager
- Buhari before He became president
- Buhari when he had a cough and sorethroat
- Buhari if he is speaking with his mouth full (but he shouldn't anyway)

And it must be able to detect it was buhari:
- when he speaks in a party with jamz rocking
- when his voice is broadcasted from a megaphone across the street
- when his recorded voice is being played with my nokia 3310
- my daddy's record discs

Under the following moods:
- happy
- sad

Under the following moods:
- crying
- singing

In the following dimensions:
- Mono
- Stereo
- 3D surround

In the following weather conditions:
- Heavy rain with winds blowing
- Thunderstorm situations


Well. . . . .if your code can fulfill all that, I see no reason why you should not win the award

- Happy coding

I've already started pulling down rouge libraries without having these pertinent cases addressed.

Youtube uses an automated approach when policing for copyright music samples in uploaded videos. There are ingenious ways around Youtube's sniffer which requires changing the music samples and/or video fidelity (youtube doesnt send minions listening to every upload for possible copyright violations and it doesnt solely count on youtube users flagging violations).

But with this project we dont have the slightest idea of what type of Buhari voice to expect, to further promise to deliver 90% accuracy of an unknown.

Somehow I started coding..

AAinEqGuinea:

I've already started pulling down rouge libraries without having these pertinent cases addressed.

Youtube uses an automated approach when policing for copyright music samples in uploaded videos. There are ingenious ways around Youtube's sniffer which requires changing the music samples and/or video fidelity (youtube doesnt send minions listening to every upload for possible copyright violations and it doesnt solely count on youtube users flagging violations).

But with this project we dont have the slightest idea of what type of Buhari voice to expect, to further promise to deliver 90% accuracy of an unknown.

Somehow I started coding..

Chai! the love of money is enough motivation. . . . .i am sure that some people have started learning programming in order to win this quiz!

dhtml18:

Chai! the love of money is enough motivation. . . . .i am sure that some people have started learning programming in order to win this quiz!

I dont need the money... but a different puzzle is always nice.

I'm having other computational issues from other projects at this moment that I'm pushing off. I'm glad to procrastinate on this (if Seun ever comes back to answer questions)... and I like see other programmer's approaches, questions, and bemoanings as well, this is good homework.

In artificial intelligence there must be a predefined training set.

Biometrics is affected by age, hence your voice, gait etc doesn't stay the same.

In this case the buhari voice should be from current speeches.

I would have attempted this assuming the code isn't open else let's see some real cash

nollyj:
Yesterday I decided to try my hand on this project. knowing that I am a beginner programmer, I am more interested on what I will learn. I did some research and read some few papers on this topic.

I decided to make a rough trial demo of 3rd Party Libraries I found. My intended language was Java.

I first wrote a desktop application using a Java library called Recognito. The link is here if you want to read more about the library.

I will say that it is a good java library for Text Independent Speaker Recognition in Java. I integrated it to my application and checked the usefulness

The NetBean Java application is attached below. I obtained the following results in four different Buhari speeches



The first two results show the positive matching result of different audio speeches from President Buhari.

The third result was 100 % positive when you match the same audio speeches from Buhari.

Finally, the fourth result is a false negative result. When I dig into to find out why the result return negative, I realize the following factors can prevent detection of a speaker voice.

1. The encoding software used to convert from any file type to .wav file

2. Noise in an audio file

3. Quality of the audio file

It is also important to understand that there is a difference between Speaker Recognition and Speech Recognition. According to wikipedia - “Speaker recognition is the identification of a person from characteristics of voices (voice biometrics). It is also called voice recognition. There is a difference between speaker recognition (recognizing who is speaking) and speech recognition (recognizing what is being said).”

I went on to test Audio /Sound Fingerprinting in android app

(The file size is big to attach here. I will look for a way to share to anyone that wants to download)

how is this different from my own submission?

I used recognito and had about the same level of accuracy.

I really don't think Seun would search for a recording of Buhari when he had soar throat or when he was crying. From the first post, it's pretty obvious almost all your stated conditions would not be required.

dhtml18:
Let me explain once and for all the kind of buhari your app must capture:
- Buhari's voice when he was a baby (crying)
- Buhari when he was a teenager
- Buhari before He became president
- Buhari when he had a cough and sorethroat
- Buhari if he is speaking with his mouth full (but he shouldn't anyway)

And it must be able to detect it was buhari:
- when he speaks in a party with jamz rocking
- when his voice is broadcasted from a megaphone across the street
- when his recorded voice is being played with my nokia 3310
- my daddy's record discs

Under the following moods:
- happy
- sad

Under the following moods:
- crying
- singing

In the following dimensions:
- Mono
- Stereo
- 3D surround

In the following weather conditions:
- Heavy rain with winds blowing
- Thunderstorm situations


Well. . . . .if your code can fulfill all that, I see no reason why you should not win the award

- Happy coding

^^^And another drama unfolds where boys are now "forking" with each others codes. . . .guys stop forking as if you are on github and start coding! E be like say seun sef don tire for the thread, i no dey see am for here again.

who made the lecturer deactivate!
dhtml18, seems u bullied d guy into deactivation.




no winner yet? the closest should get the reward then.

Seun is there any time frame for this competition? Or is it until someone meets up to 90% accuracy?

Can I make another submission? I guess I should be able to.

90% is feasible. I think.

ayobase:
who made the lecturer deactivate!

dhtml18, seems u bullied d guy into deactivation.





no winner yet?
the closest should get the reward then.

No mind me jare - I just wanted to show power small say me sef fit write Epistle if I like.

5 days and still counting, no one seems to be capable here o. I am too busy or i would have made an attempt. Kudos to all those attempting to win the award.