speaker recognition in matlab

[hansmuller]

speaker recognition matlab

Hi all.

What is the parameter that compares two or more speakers speech? I mean what is the characteristic of a person's voice.

I want to make speaker recognition in matlab. I want to decide which person is speaking regardless of the word.

Can anyone help me please ?

Thanks in advance.

 

[badboyz]

speaker recognition using matlab

Try computing Euclidean distance between feature extracted parameters from speakers.

 

[hansmuller]

speaker recognition

Thanks for your reply.

I am already wondering these features. I mean feature extraction must be done by MFCCs or any other techniques? Or just frequency info is enough for comparison?

 

[supiper]

speaker recognization matlab

Hi,
Many features can be used to compare two or more speaker’s speech. The most used is MFCC (generally 12 coefficients and 2 dynamics:delta and delta-delta). You can also extract LPC (Linear Prediction Coefficient, PLP (Perceptual Linear Prediction, RASTA-PLP,...), or you can simply use Fundamental Frequency and/or Format frequencies...

 

[hansmuller]

matlab speaker recognition

Hi supiper,

Thanks for your reply. Actually I am using MFCCs for recognition of words for same speaker. It is useful. But how will I use the coefficients for recognition of speakers regardless of the spoken words.

Can I use only fundamental frequencies to make a comparison? Does it work?

Thanks.

 

[badboyz]

speaker recognition from edaboard.com

After MFCC you can go for vector quantization (Compute the euclidean distance) fix threshold based on the max. euclidean distance and detect the word.

 

[hansmuller]

perceptual linear prediction in matlab

Yes badboyz, youre right. I am already using euclidean distance to detect same words. But here I am looking for speakers not words.

 

[badboyz]

vector quantization in speaker recognition matlab

Ok you can try this threshold thing on different speakers also.

 

[cross de lena]

speaker identification matlab

What you are speaking is a speaker verification system,as badboyz said,in addition to all what u do, a speaker-specific threshold is also computed from the trained samples. During the testing (operational) phase the input speech is matched with stored reference model(s) and recognition decision is made. But this will be challenging coz of the highly variant nature of input speech signals.Speech signals in training and testing sessions can be greatly different due to many facts such as people voice change with time, health conditions (e.g. the speaker has a cold), speaking rates, etc.Also variations in recording environments plays a major role.

 

[supiper]

matlab euclidean speech

So you speak about “Speaker Identification” (it's not the same thing with word recognition). Then, after coefficient computation you can use the DTW algorithm to search similarity. See this link for more explanation about the DTW algorithm:

https://en.wikipedia.org/wiki/Dynamic_time_warping

 

[hansmuller]

mfcc matlab algorithm

yes i am working on speaker identification. I have done word recognition with MFCCs and some use of DTW as you told. In the same way can i recognize speakers? in the same algorithm ?

 

[supiper]

speaker recognize matlab programs

Hi,

See this two files for more details about speaker identification/recognition!

 

[hansmuller]

speaker identification matlab code

Supiper, thanks for documents. I think they will help me.

 

[hansmuller]

speeker recognition matlab

Dear friends,

For isolated word recognition I am creating MFCCs from the filtered and endpoints detected voice. After created MFCCs I look the distances between created MFCCs of all words. Then the minimum distance is the recognized word.

Is there anything wrong or missing by this system? Should I do anything after creating MFCC vectors?

Thanks in advance.

 

[bhooms]

hi i am new in matlab and dsp........
in the preprocessing step for speaker recognition using MFCC and VQ
whether we have to convert wav sound file into digital format and if yes how.....
please do help i m really confused

 

[radicalryan]

hi, im also working on speaker recognition..Ahm, i just want to ask this, how can i apply euclidean distance formula on LPC coefficients? My output on LPC is an array of coefficients which the features extracted..how can i detemine which of these coefficients should i apply the euclidean provide that my input audio is different from those in the database? should you first arrange these coefficients before you use the euclidean distance.

hope for reply..

 

[midex87]

Hey, i am also doing a speaker verification program in matlab.

I know you need to extract features using an algorithm like mfcc for and vq for modelling a codebook for the speaker (also referred to classification).

What im not sure of is how to compare both speaker 1 with speaker 2 and determine if it is the same speaker.

Eucildean Distances computes the distortion between both speaker 1 and speaker 2s codebooks but how do i set a threshold value????


Any assistance would be appreciated.

 

[sqiey]

hi...i'm doing the speaker verification using matlab...can you help me with giving the full code because i'm only can record the voice but don't know how to verify it using the database...i'm really new in matlab....

 

[midex87]

read this paper, it might help, it helped me!

**broken link removed**

 

[sqiey]

this coding only to recognize word or can verify the speaker??

Added after 3 hours 45 minutes:

can anyone give me an example of electronic circuit for speaker verification??