Heart sound identification refers to a technology that uses the human heart sound signal to identify the heart. The heart sound is a reflection of the mechanical movement of the heart and cardiovascular system, including the physiological and pathological information of the various parts of the heart and their interactions. Therefore, the heart sound signal has completely different characteristics in different people and has high stability, which can be used as an identification feature of biometric technology. In addition to being difficult to disguise, the heart sound signal has the advantage of being easy to obtain in addition to forgery and tampering. Therefore, the human heart sound signal can be a novel biometric method.
This paper develops an embedded heart sound identification system based on LabVIEW. The system is convenient and flexible, and can realize the registration, identification and confirmation of user identity.
1, heart sound signal identification principle
The heart sound signal used for personal identification mainly includes two steps: feature extraction and pattern matching. Feature extraction is an effective and stable and reliable feature that extracts the heart sound signal and uniquely represents the identity of the subject. Pattern matching is for training and authentication. The feature pattern is similarity matching. In this paper, the heart sound identification system is designed based on the recognition algorithm of MelFrequency Cepstrum Coefficient (MFCC) feature extraction and vector quantization (VQ) model matching.
(Image from the network)
The MFCC is mainly used for speaker recognition, which converts the spectrum into a frequency-based nonlinear spectrum and then converts it to the cepstrum.
Appropriate improvements to the MFCC can be applied to the identification of heart sounds. According to the frequency domain characteristics of the heart sound signal, the cutoff frequency of the Mel filter group is selected to be 500 Hz; since the heart sound signal has quasi-periodicity and no strong non-stationarity of the speech signal, the frame length of the signal is selected to be 256 ms instead of 20 ms of the speech signal. The MFCC coefficient is selected to be 32 orders, and the high-order MFCC coefficients of the heart sound signal contain more information, and the first-order difference coefficient can make the signal dynamic characteristics strong. VQ is derived from the promotion and development of scalar quantization. Scalar quantization is to use discrete numerical values ​​to represent discrete time-domain signals with continuous values ​​for each amplitude. Vector quantization is to divide several time-domain sampling signals with continuous amplitude values ​​into a group, that is, form a vector, and then A number of discrete digital values ​​are used to represent the various vectors. In the study of pattern recognition, the task of classifying each vector to be identified needs to be completed. The VQ-based heart sound identification model is simpler and more real-time than other recognition models (such as Gaussian mixture models).
2, system implementation 2.1 hardware system implementation
The hardware of this system is composed of upper computer and lower computer. The overall structure is shown in Figure 2. The HC-06 Bluetooth module communicates with the HC-06 Bluetooth module. The HC-06 Bluetooth module uses CSRBC04 Bluetooth technology, built-in Bluetooth antenna, and the transmission power is Class2, and the sensitivity can reach -80dBm.
The lower computer uses the dsPIC digital signal processor as the core to control the acquisition, amplification and ADC of the heart sound signal, and then sends the heart sound signal to the upper computer through the HC-06 Bluetooth module (slave). The lower computer mainly includes the analog circuit and the digital circuit. section. Analog circuits include heart-tone sensors, 30Hz high-pass, 500Hz low-pass Butterworth filters, and gain-adjustable audio amplifiers. The heart sound sensor consists of a stethoscope probe, an electret microphone, and a catheter; the Butterworth filter is a 4th order, and the Sallen-Key structure is selected, which allows independent gain setting. The gain adjustable audio amplifier uses LM4811, and its CLOCK and UP/DN pins are connected to the RG6 and RG7 pins of the dsPIC to realize the control of the heart sound signal amplification. The digital circuit is mainly dsPIC main control chip and HC-06 bluetooth module. The dsPIC33FJ128MC506 is selected as the main control chip. The system clock is set to 40MHz, the sampling frequency is set to 2kHz, the baud rate is set to 11.5kbps, and the analog signal is passed through its 12-bit ADC. After being converted to a digital signal, it is transmitted by the UART to the HC-06 Bluetooth module (slave).
The upper computer is a terminal with the industrial control board as the core, and displays, analyzes, stores and recognizes the heart sound signal. Including HC-06 Bluetooth module (main), TTL-RS232 level conversion circuit, industrial control board and LCD touch screen display, HC-06 Bluetooth module (main) receiving and transmitting heart sound signal through the level conversion circuit to change the TTL level RS232 level; industrial control board connects to the Bluetooth module (main) through RS232 interface to complete the reception of heart sound signals; LCD touch screen display as human-computer interaction device, connected to the industrial control board through LVDS bus and USB bus, LVDS bus transmits video signal The USB bus transmits touch signals.
2.2 Software System Implementation
The self-designed WindowsEmbeddedStandard operating system is chosen. Because it is a componentized XP system, it does not need to design file system and develop drivers, which greatly shortens the development cycle and development difficulty. Unique enhanced write filter technology reroutes selected disk I/O to internal memory or other storage media, allowing the operating system to think that your read-only memory is writable; custom boot screens and custom shells can be customized The running shell of the system starts to prevent malicious modification of system configuration or misoperation to damage the system, ensuring platform stability and data security.
3, heart sound identification software development
This paper uses LabVIEW virtual instrument to develop heart sound identification software based on MFCC feature extraction and VQ pattern matching algorithm. The software design process needs to use NI DatabaseConnectivityToolkit toolkit. DatabaseConnectivityToolkit provides complete SQL function, using MicrosoftADO technology and most commonly used databases. Connect to enable interactive operations with local or remote databases. The heart sound identification software implements three functions: user registration, user identification and user confirmation. Each function has an independent function interface, and the Subpanel is used to implement the dynamic loading interface.
User registration is divided into two steps: the first step is the input storage of basic information, including user name, age, gender, etc. The second step is to collect heart sound signals, extract their MFCC feature parameters, and use LBG algorithm to generate an optimal codebook storage. In the local database. The user identification is a 1:N mode, collecting the heart sound signal of the user to be identified, extracting the MFCC feature parameters, comparing with all the user codebooks already existing in the local database, and selecting the optimal codebook according to the minimum average quantization distortion degree criterion. Make a match. The user confirms that it is a 1:1 mode. First, the user inputs the registered user name, and then collects the heart sound signal of the user to be identified, extracts the MFCC feature parameter, and compares it with the codebook of the user-specified identity, if the average quantization distortion is If it is less than the established threshold, its identity is confirmed.
In the future, the heart sound denoising scheme based on adaptive enhancement technology will be studied to improve the robustness of the recognition algorithm. Although the EER is 6.67% in the case of this small-capacity template experiment, in the case of large-capacity templates, in order to ensure that EER is still For a small value, the threshold selection in the user confirmation mode will be the focus of the next step.
4 Conclusion
Based on the principle of MFCC feature extraction and VQ pattern matching recognition, this paper develops an embedded heart sound identification system based on LabVIEW, which has user registration, user identification and user confirmation. The high CRR and lower EER under the small-capacity heart sound template fully prove the feasibility of the embedded heart-tone identification technology, which will provide a high accuracy rate for various identification and information security problems faced by the society. New biometric identification equipment with strong anti-counterfeiting capabilities.
This paper develops an embedded heart sound identification system based on LabVIEW. The system is convenient and flexible, and can realize the registration, identification and confirmation of user identity.
1, heart sound signal identification principle
The heart sound signal used for personal identification mainly includes two steps: feature extraction and pattern matching. Feature extraction is an effective and stable and reliable feature that extracts the heart sound signal and uniquely represents the identity of the subject. Pattern matching is for training and authentication. The feature pattern is similarity matching. In this paper, the heart sound identification system is designed based on the recognition algorithm of MelFrequency Cepstrum Coefficient (MFCC) feature extraction and vector quantization (VQ) model matching.
(Image from the network)
Appropriate improvements to the MFCC can be applied to the identification of heart sounds. According to the frequency domain characteristics of the heart sound signal, the cutoff frequency of the Mel filter group is selected to be 500 Hz; since the heart sound signal has quasi-periodicity and no strong non-stationarity of the speech signal, the frame length of the signal is selected to be 256 ms instead of 20 ms of the speech signal. The MFCC coefficient is selected to be 32 orders, and the high-order MFCC coefficients of the heart sound signal contain more information, and the first-order difference coefficient can make the signal dynamic characteristics strong. VQ is derived from the promotion and development of scalar quantization. Scalar quantization is to use discrete numerical values ​​to represent discrete time-domain signals with continuous values ​​for each amplitude. Vector quantization is to divide several time-domain sampling signals with continuous amplitude values ​​into a group, that is, form a vector, and then A number of discrete digital values ​​are used to represent the various vectors. In the study of pattern recognition, the task of classifying each vector to be identified needs to be completed. The VQ-based heart sound identification model is simpler and more real-time than other recognition models (such as Gaussian mixture models).
2, system implementation 2.1 hardware system implementation
The hardware of this system is composed of upper computer and lower computer. The overall structure is shown in Figure 2. The HC-06 Bluetooth module communicates with the HC-06 Bluetooth module. The HC-06 Bluetooth module uses CSRBC04 Bluetooth technology, built-in Bluetooth antenna, and the transmission power is Class2, and the sensitivity can reach -80dBm.
The lower computer uses the dsPIC digital signal processor as the core to control the acquisition, amplification and ADC of the heart sound signal, and then sends the heart sound signal to the upper computer through the HC-06 Bluetooth module (slave). The lower computer mainly includes the analog circuit and the digital circuit. section. Analog circuits include heart-tone sensors, 30Hz high-pass, 500Hz low-pass Butterworth filters, and gain-adjustable audio amplifiers. The heart sound sensor consists of a stethoscope probe, an electret microphone, and a catheter; the Butterworth filter is a 4th order, and the Sallen-Key structure is selected, which allows independent gain setting. The gain adjustable audio amplifier uses LM4811, and its CLOCK and UP/DN pins are connected to the RG6 and RG7 pins of the dsPIC to realize the control of the heart sound signal amplification. The digital circuit is mainly dsPIC main control chip and HC-06 bluetooth module. The dsPIC33FJ128MC506 is selected as the main control chip. The system clock is set to 40MHz, the sampling frequency is set to 2kHz, the baud rate is set to 11.5kbps, and the analog signal is passed through its 12-bit ADC. After being converted to a digital signal, it is transmitted by the UART to the HC-06 Bluetooth module (slave).
The upper computer is a terminal with the industrial control board as the core, and displays, analyzes, stores and recognizes the heart sound signal. Including HC-06 Bluetooth module (main), TTL-RS232 level conversion circuit, industrial control board and LCD touch screen display, HC-06 Bluetooth module (main) receiving and transmitting heart sound signal through the level conversion circuit to change the TTL level RS232 level; industrial control board connects to the Bluetooth module (main) through RS232 interface to complete the reception of heart sound signals; LCD touch screen display as human-computer interaction device, connected to the industrial control board through LVDS bus and USB bus, LVDS bus transmits video signal The USB bus transmits touch signals.
2.2 Software System Implementation
The self-designed WindowsEmbeddedStandard operating system is chosen. Because it is a componentized XP system, it does not need to design file system and develop drivers, which greatly shortens the development cycle and development difficulty. Unique enhanced write filter technology reroutes selected disk I/O to internal memory or other storage media, allowing the operating system to think that your read-only memory is writable; custom boot screens and custom shells can be customized The running shell of the system starts to prevent malicious modification of system configuration or misoperation to damage the system, ensuring platform stability and data security.
3, heart sound identification software development
This paper uses LabVIEW virtual instrument to develop heart sound identification software based on MFCC feature extraction and VQ pattern matching algorithm. The software design process needs to use NI DatabaseConnectivityToolkit toolkit. DatabaseConnectivityToolkit provides complete SQL function, using MicrosoftADO technology and most commonly used databases. Connect to enable interactive operations with local or remote databases. The heart sound identification software implements three functions: user registration, user identification and user confirmation. Each function has an independent function interface, and the Subpanel is used to implement the dynamic loading interface.
User registration is divided into two steps: the first step is the input storage of basic information, including user name, age, gender, etc. The second step is to collect heart sound signals, extract their MFCC feature parameters, and use LBG algorithm to generate an optimal codebook storage. In the local database. The user identification is a 1:N mode, collecting the heart sound signal of the user to be identified, extracting the MFCC feature parameters, comparing with all the user codebooks already existing in the local database, and selecting the optimal codebook according to the minimum average quantization distortion degree criterion. Make a match. The user confirms that it is a 1:1 mode. First, the user inputs the registered user name, and then collects the heart sound signal of the user to be identified, extracts the MFCC feature parameter, and compares it with the codebook of the user-specified identity, if the average quantization distortion is If it is less than the established threshold, its identity is confirmed.
In the future, the heart sound denoising scheme based on adaptive enhancement technology will be studied to improve the robustness of the recognition algorithm. Although the EER is 6.67% in the case of this small-capacity template experiment, in the case of large-capacity templates, in order to ensure that EER is still For a small value, the threshold selection in the user confirmation mode will be the focus of the next step.
4 Conclusion
Based on the principle of MFCC feature extraction and VQ pattern matching recognition, this paper develops an embedded heart sound identification system based on LabVIEW, which has user registration, user identification and user confirmation. The high CRR and lower EER under the small-capacity heart sound template fully prove the feasibility of the embedded heart-tone identification technology, which will provide a high accuracy rate for various identification and information security problems faced by the society. New biometric identification equipment with strong anti-counterfeiting capabilities.
It is a bit that, when rotated, drills a hole larger than its diameter.
It is including Matrix body and Steel body material .
The Bi-Center Bits provide cost-effective hole opening through installed casing or liner sections.
Engineered cutting structure designed .
High performance cutters from Great factory .
Gauge cutters designed Arc PDC cutters ,which is more abrasive then normal hard alloy gage pads .
Longer working life to save your drilling cost .
Bi-Center Bits,Bi-center PDC drill bits, Matrix body Bi-center bits,Bi Center PDC Bits
CANGZHOU GREAT DRILL BITS CO., LTD , https://www.pdcbits.nl