Design Of Voiceprint Cryptographic Lock Application Inandroid System

As a major mobile-phone operating system, Android smart-phone mainlydepends on focus activation, hand cranking control and password for system security,these modes are easy and quick, but not unique, others can get certified and gainrights of cell phone usage once they steal the password information, resulting ininformation disclosure. Therefore, this paper proposes an authenticate model basedon voiceprint characteristics, and builds a voiceprint lock with double protectionmechanism by collecting the owner unique voiceprint features and semanticinformation, while apply it for boot, sleep start or privacy file protection andsoftware usage could implement authorization usage of cell phone and preventleakage of sensitive personal data, thence heighten the safety usage level of phone.Firstly, make a comparative analysis on linear predictive cepstral (LPCC) andMel frequency cepstral coefficients (MFCC), study their impact on the falserejection rate(FRR) and false acceptance rate(FAR), extract an unique characteristicby experiment. Then consider Android cell phone’s computing power limitations andthe balance on recognition rate(RR) and recognition time(RT), compare and analysisthe text-related voiceprint recognition effect of dynamic time warping (DTW) andvector quantization (VQ) on the condition of small samples, experiment shows thatVQ’s RT is only0.16s,0.3s faster than DTW, while its RR can achieve90.5%and itsFRR, FAR are both lower than DTW. To enhance the robustness of algorithmsapplied in noise environment, study the speech enhancement and endpoint detectiontechniques, propose an improved threshold function for lack of hard and softthreshold function in wavelet threshold de-noising, and have a comparative analysison the improved method’s noise filtering effect; In addition, propose a combinedvoice activity detection(CVAD) method based on fractal dimension and energy,experiment proves that it can still maintain75%of endpoint detectionaccuracy(EDAA) even in the environment of-5dB and can reach more than95%in ahigh SNR environment.Finally, implement a voiceprint lock application in Android system, the ownerhimself can always change the password in order to avoid the problem of information leakage and recording playback. It can be achieved more than90percentRR in lab environment by experiment, and its FRR, FAR are respectively reduced by17.5%and20%while its RR is from61%up to88.5%, after the adoption ofimproved wavelet threshold method and CVAD technology, effectively enhance thealgorithms robustness of complex application environment. It can be installed andrun on all Android devices, otherwise, it can select the sampling rate according todifferent versions of Android devices to record voice signal and extract speech data,and the owner can change password anytime to prevent password disclosure. Themost important is that the RR and RT are within the acceptable range.