numberc / gset-ai-authentication-a Goto Github PK

GSET MATLAB FILES

FOLDERS

• classes: contains data structures to organize data
• functions: contains functions to calculate signal information
• recordings: contains audio files from user authentication experiments, subfolders for individual users
• transmissions: contains chirp audio files transmitted by mobile device
• user_data: contains matlab variable files storing features for each user

CLASSES

• profile.m: structure for user data, holds user-influenced chirp signal and features extracted

FUNCTIONS

• func_chirp_gen.m: generates chirp signal based on parameters provided to it
• get_features: calculates time domain features when given user-influenced chirp signal
• get_freq_data: calculates frequency domain features such as mfcc, called by get_features
• mfcc.m: calculates mfcc variables
• trifbank.m: used by mfcc.m
• vec2frames.m: used by mfcc.m

MATLAB PROGRAMS <--- THIS IS THE REALLY IMPORTANT STUFF

• Step 1: create_chirp.m: choose what type of chirp signal to create, saves .wav file
• Step 2: cross_correlate.m: select audio file from audio folder and extract user-influenced signal from recording, creates profile ** Step 2 (Alternate): thresholding.m: cross_correlate may sometimes incorrectly calculate the user signal, thresholding is simpler method
• Step 3: feature_extraction.m: given user profile, calculate the features, saves user data into profile
• Step 4: learning.m: compiles all desired profiles into one data structure for classification learner app
• Step 4: gesture.m: compiles gesture data to test a different approach

CLASSIFICATION LEARNER APP <--- ALSO REALLY IMPORTANT

• In MATLAB, open the APPS tab at the top of the screen
• Click the Classification Learner icon (looks like bunch of red and blue dots)
• Start a new learning session by loading your desired workspace data (master_data created by learning.m)