Cancelable Speaker Identification Based on Inverse Filter
Marwa A. Elsayed, Walid El‐Shafai, Mohsen Rashwan, Moawad Ibrahim Dessouky, Adel S. El‐Fishawy, Fathi E. Abd El‐Samie · Menoufia Journal of Electronic Engineering Research · 2019
Nowadays, biometric systems have replaced the password or token based authentication systems in many fields to improve the security level. Biometric authentication systems automatically identify or verify a person using physical, biological, and behavioral characteristics. Unfortunately, these systems have suffered from hacking trials. This paper presents masking techniques for cancelable speaker identification. The basic concept is to use another version of the original speaker signal created through a masking techniques to keep the original signals safe and away from utilization in the system. The basic idea is adding some sort of noise to the speech signal and magnification of this noise by the proposed techniques. The proposed techniques are discrete wavelet transform (DWT), wavelet thresholding, Linear Minimum Mean Square Error (LMMSE) technique and invers filter. A comparison between these techniques is established using objective quality measures of speaker identification. Simulation results show good performance of the technique based on invers filter.