Send the link below via email or IMCopy
Present to your audienceStart remote presentation
- Invited audience members will follow you as you navigate and present
- People invited to a presentation do not need a Prezi account
- This link expires 10 minutes after you close the presentation
- A maximum of 30 users can follow your presentation
- Learn more about this feature in our knowledge base article
Do you really want to delete this prezi?
Neither you, nor the coeditors you shared it with will be able to recover it again.
Make your likes visible on Facebook?
You can change this under Settings & Account at any time.
Final Year Project: Voice Recognition system using MATLAB
Transcript of Final Year Project: Voice Recognition system using MATLAB
Yusherizan Marshella Binti Yusoh
(EE084076) BACKGROUND CONCLUSION Steps to construct the Voice Recognition System:
Prepare a speech database for training and testing.
Simulation and evaluation. RESULT AND ANALYSIS OBJECTIVES Construct a Voice Recognition System using MATLAB.
Create a MATLAB coding to put together a useable system.
Coding created using MFCC extraction technique and VQ model. Voice Database for Training and Testing Create voice database for references to make comparison later on.
Voice database recorded using Microsoft WAV format.
Create database for the TRAIN folder and TEST folder.
The voice model(VQ codebook) will be trained in MFCC vector space for each speaker.
After the training steps, system should be able to acknowledge the voice characteristic for each speaker.
Testing phase, system will identify the speaker of each sound file in TEST folder. Voice Processing Transform the sound file to MFCC sequence (acoustic vectors).
As example, when running the sound file from train folder (train1) this result will be obtain:
-Sampling rate: 12500
-Highest frequency: 12500Hz
-Msec in 256 samples= 20.48msec
Signal will be plotted in time domain.
Contain high amount of data causing problem for examining the voice.
Speech signal(vector) will be cut into frames Voice Processing The speech signal(vector) will be cut into frames. Result of cutting will be a matrix of N samples from original voice.
Applying windowing and FFT to transform the signal into frequency domain (spectrum or periodogram)
Compute power spectrum using imagesc command.
Converting power spectrum into MFCC. Voice Processing VQ is needed to build a speaker references models from vectors in training phase (obtained from MFCC) and to identified any sequence of acoustic vector uttered by unknown speakers.
Inspection on vectors is done by choosing any two dimensions and plot the data points.(eg:5th and the 6th)
Observation needed to see whether data region for two different spaeker overlapping each other and in cluster.
Function vqlbg will train VQ codebook using LGB algorithm. VECTOR QUANTIZATION (VQ) SIMULATION AND EVALUATION Simulate the training and testing procedure for the voice recognition system. Technique chosen for Feature Extraction and Feature Matching Feature Extraction:
Mel Frequency Ceptrum Coefficient (MFCC)
Vector Quantization (VQ) METHODOLOGY Voice Recognition System: Process of automatically recognizing who is the spaeker based on the unique characteristics contained in speech waves.
System implemented as security features for:
-Credit card Verification
Two elemets in Voice Recognition System:
- Feature Extraction: Process of extracting
unique information from speech files.
- Feature Matching: Process of identifying the
speaker that involves comparing unknown
data. Supervisor: Mr Wong Hung Way
Examiner: Prof. Dr Madya Abu Bakar bin Mhd Ghazali OVERVIEW INTRODUCTION
RESULT AND ANALYSIS
CONCLUSION Improvement can be implement in the Voice Recognition System according to the application.
Application that utilize the Voice recognition system:
-Time and attendance system
- Biometric Login to telephone
aided shopping systems