Oral Presentation , Page 288-294 (7)

Volume Title: 2nd IEEE International Conference on Electronic Eng., Faculty of Electronic Eng., Menouf, Egypt, 3-4 July. 2021

¹mona kareem shafeek ; ²Adel Abdel-masseh Saleeb; ³Adel shaker el fishawy; ⁴Moawad Dessouky; ⁵sami el dolil; ⁶Fathi Abd El-SAmie

¹Communications and Electronics Department Faculty of Electronic Engineering, Menoufia University: Menouf, Egypt

²Faculty of Electronic Engineering, Elgesh Elbahry Street

³Department of Electronics and Electrical Communications Engineering Faculty of Electronic Engineering Menoufia University, Egypt

⁴Communications and Electronics Department Faculty of Electronic Engineering, Menoufia University: Menouf, Egypt,

⁵Communications and Electronics Department Faculty of Electronic Engineering, Menoufia University: Menouf, Egypt

⁶Minufia- Egypt

Among all types of biometric recognition technologies, speaker identification has been considered as one of the most popular technologies due to simplicity and ability to capture and encode speech signals. However, biometric features may be misused by third parties. Hence, these features should be secured through the generation of revocable biometric templates representing individuals. This paper is concerned with cancelable speaker identification systems. It presents a proposed cancelable biometric recognition technique to enhance the security in addition to the classification accuracy of the speaker identification process. This technique relies on comb-based digital filters to intentionally deform speech signals for cancelable template generation. Zero cancelation is adopted to create low-pass and high-pass versions from the comb filter for speech signal deformation. The proposed digital filtering technique has the advantage of irreversibility, because the filter frequency response has multiple bands with multiple nulls. This induces one-way non-invertible deformations as a result of the nulling effect [1]. Hence, this proposed technique achieves an adequate security level in comparison with traditional unsecured speaker identification techniques. Simulation results reflect the acceptable classification accuracy and security in spite of the deterioration effect due to filtering.

Speech processing; comb filter; and MFCCs