Design Of Audio Key System Based On USB Audio Device Class

With its versatility and convenience in use, USB has become a growing number of handheld devices and even larger external equipment's standard interface. USB interface for transmission in the key security system is also universal, but simply to stored data in the USB key can not guarantee the security of user data, the loss of key data may bring user information theft.Speech recognition technology can date from 1950s and 1960s ,and it has entered a more mature stage of the application, so this paper considers the use of the article that user can never lose: the voice as the security guarantee. The system chooses to strike the MFCC coefficient of a section of voice as the save for users validation, In addition, raises a design that achieves the acquisition and processing of voice signal in an embedded handheld device. This will contribute to a separate encryption to the characteristics of the voice data in the equipment side to further enhance the security of the system; also alleviate the pressure on the server-side computing, to make the remote server verify key become possible.In this paper, USB interface is used as the information exchange interface between the key equipment and mainframe server-side. Based on the USB Audio equipment protocol, the article designs a USB audio device structure to control the microphone. In order to enhance system versatility, it chooses Windows xp as the server operating system. It make a design of the USB key system communication protocol between device-side and host-side, for the host to configurate ADC sampling features at audio equipment side, and the program of value of MFCC characteristics extraction. Meanwhile,for the equipment to feedback the configuration of the current audio ADC and sent the MFCC characteristics already got to host-side. This set of communication protocols is built on the standard operating system drivers usbaudio.sys ,which is realized by calling the Windows api audio library function on the host side. So the system doesn't need to develop or install special host drivers only , but also simplifies the use of the system.This paper analyzes the MFCC feature extraction algorithm. It gives a method that is , getting a certain period of time of admission to frame the environmental voice, in order to do improved spectral subtraction operation to the admission to voice later through analysis to the noise spectrum, in this method white noise is eliminated. For the residual "music noise", it gives a design of the " Method of Mask Comparing with First Time Recorded Key Voice ", that is, in the host side, do improved spectral subtraction operation to the first time recorded key voice, and then achieve its masking threshold for preservation.Then every time when doing verification operation, just comparing with environmental noise power spectral density with the masking threshold to determine whether to do masking noise operation. This method reduces the complexity of the algorithm while enhancing the accuracy of the algorithmThe paper improves floating-point to fixed-point arithmetic structure for the embedded system environment. In accordance with the MFCC algorithm steps,this paper classifies fixed to floating-point conversion nto three categories: simple floating-point parameters conversion, simple floating-point function conversion and specific floating-point function conversion. For simple floating-point function conversion,it adepts coordinate rotation digital computer (CORDIC) algorithm. For specific floating-point function conversion,it makes use of look-up table method. In order to reduce the size of the device-side firmware,it makes a pre-establishment and preservation of the list of specific floating-point function value on the host side. Before voice authentication operation, through the custom USB Key audio communication protocols, function value list is sent to the USB device in accordance with user's configuration of the audio equipment and MFCC feature extraction on the host side.This improvement greatly reduces the space occupied by the firmware.This article attempts to combine the universality of USB interface and security of speech recognition technology, , to achieve a safe and convenient hand-held Audio-Key, and completes the design from device-side firmware to host-side application. The experimental test results show that,for around 3 seconds long time secret voice under the highest sampling rate, 16K Hz, this system can guarantee both real-time and the recognition rate of 82.5 percent. The proposed program, the Audio-Key communication protocols and improved algorithms have reached the desired results.