Speaker Verification System in a Security Application HŪDATBrian Bash Thomas Jonell Dustin Williams Advisor Dr. Les Thede
Executive Summary Utilize voice verification in a security application Interface with an electronic lock Be cost advantageous compared to other solutions
Points of Interest Problem Identification What are Biometrics? Constraints Project Layout Cost Analysis Gantt Chart Conclusion
Problem Identification What? Desire to control access to a secure room by means other than a traditional keycard or passcode Why? Traditional methods can be compromised
Biometrics Growing Field of Technology Identification of individuals Uses biological traits Examples: Retinal or Iris Scanning Fingerprints Facial Recognition
Applicable Biometrics Facial Recognition Time consuming Difficult to implement Accurate Voice Verification Quick Easier implementation Not as accurate
Voice Biometrics Speech Recognition Program as a stenographer Speech to text programs Speaker Verification Used in security Does not care what you say Concerned about the nature of your voice
Constraints Economic Utilize standard parts Limiting software developing packages Keep price competitive Manufacturability Utilize standard parts Sustainability Proper documentation Sufficiently commented code
Constraints cont. Health and Safety Debug for all feasible scenarios Manual override Political / Ethical Future – abuse of collected bio-data General scope: no concerns Environmental and Social No foreseen impact
Design Computer-driven Microphone Audio Processing Program Locking Mechanism
Block Diagram Interface / Power Supply Microphone Computer ID Electric Door Lock User System Admin.
Programming Language MATLAB A very diverse program written in C Many built in functions Intuitive interface Digital Signal Processing (DSP) applications
Voice Verification Program Voice Recorder Voiced/Unvoiced Detection Voice Analysis Testing
Voice Recorder Sampled at 11025Hz for 5 seconds Stored in WAV format Naming Convention Lower case letters First initial followed by last name Ex: tjonell.wav
Voiced/Unvoiced Detection Enhance voice sample with autocorrelation. Detect spoken portions using enhanced voice sample. Split voice sample into spoken portions only.
tjonell.wav with Autocorrelation
Voice Analysis Power Spectral Density (PSD) The concentration of power at specific frequencies. Formants Characteristic resonant region in the PSD of a sound.
Voice Testing Compare the relative locations of formants in each word. A weighted score is created. This is based upon how closely a person’s voice matches their reference sample.
bbash vs. brian2
bbash vs. tjonell
Lock Circuit
Text-based User Interface ******************************* * * * Welcome to the * * Hudat Security System * * * ******************************* Please begin by entering your user name: bbash When you are ready to begin, press any key and recite the passphrase slowly into the microphone. PROCESSING... Access granted. Welcome.
Cost Analysis
Gantt Chart
Conclusion Biometrics answer increased demand for security. Voice verification is very practical. MATLAB is ideal for this application. Lock circuit allows communication between program and a door lock.
Question & Answer Session