VocalLock: Sensing Vocal Tract for Passphrase-Independent User Authentication Leveraging ...
summary
A smartphone system senses the vocal tract with FMCW acoustic sensing (20 ms period, 16–24 kHz) to capture static shape and dynamic movement, converts frequency-difference traces to MFCC-like features via an encoder–decoder feature-transfer network, and trains a GMM-UBM classifier. The prototype achieves about 91% accuracy, false-reject rates below 2%, and 82.7% one-utterance logins, resists replay and mimic spoofing, and runs passphrase-independently.
AI-written summary · about
chapters (8)
- 0:00Start
- 0:28Background
- 0:54Existing Approaches
- 1:36Passphrase-independent Authentication
- 2:31Preliminary How to implement?
- 3:09Sensing Vocal Tract with FMCW Signal Design & Processing • Period: time-invariant behavior within 20ms • Bandwidth: sensing resolution + device limitation and
- 4:47Overall Performance
- 5:23Conclusion Observation
related talks
06mVoice In Ear: Spoofing-Resistant and Passphrase-Independent Body Sound Authentication
33mYour voice confirms my identity
35mThe UX Of Advanced Voice Biometrics For Authentication
19mThumprint: Socially-Inclusive Local Group Authentication Through Shared Secret Knocks
06mEarDynamic: An Ear Canal Deformation Based Continuous User Authentication Using In-Ear Wearables
51mCan You Hear Me Now? Security Implications of Voice as the New Keyboard
05mFine-Grained and Context-Aware Behavioral Biometrics for Pattern Lock on Smartphones
06mAnti-Spoofing Voice Commands: A Generic Wireless Assisted Design