An Acoustic Feature Extraction Method for Recording Playback Attack Detection in Voiceprint Recognition
A technology of playback attack and voiceprint recognition, which is applied in the field of acoustic feature extraction for recording playback attack detection in voiceprint recognition, can solve problems such as insufficiency, and achieve the effects of improving performance, shortening time, and strengthening channel differences.
Active Publication Date: 2021-07-13
上海企创信息科技有限公司
View PDF5 Cites 0 Cited by
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
In fact, compared with the original voice, the recording and playback attack voice has two additional processes of recording and playback, and the frequency response characteristics of the recording device and the playback device are non-uniform, so that the frequency spectrum will be different in the low frequency band and the high frequency band. Therefore, only emphasizing the spectral information in the low frequency band is not sufficient
Method used
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View moreImage
Smart Image Click on the blue labels to locate them in the text.
Smart ImageViewing Examples
Examples
Experimental program
Comparison scheme
Effect test
Embodiment
[0055] Such as figure 1 As shown, it is an acoustic feature extraction method for recording playback attack detection in voiceprint recognition in an embodiment of the present invention, and the method includes the following steps:
[0056] S10. Preprocessing the input voice;
[0057] The preprocessing in step S10 includes framing, windowing and denoising, and step S10 specifically includes the following steps:
[0058] S11, the input voice signal x (n) is divided into frames, the voice signal is divided into a plurality of voice frames with a frame length of N (actually optional 1024), there is overlap between adjacent two frames, and the frame shift is L (actual optional256);
[0059] S12, adding a window to each frame of voice signal x(i, n) after framing, multiplying each frame of voice signal by a Hamming window with a window length of N, to obtain a windowed voice frame Calculated as follows:
[0060]
[0061] S13. Calculate the short-term energy SE(i) of each fr...
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More PUM
Login to View More Abstract
The invention discloses an acoustic feature extraction method for recording and playback attack detection in voiceprint recognition. The method includes: preprocessing the input voice; performing Fourier transform on each frame of the preprocessed voice signal; Convert the signal into a frequency domain signal to obtain the spectrum of the speech signal; calculate the spectral line energy of each frame of speech signal after Fourier transform; perform asymmetric hyperbolic sinusoidal frequency scale transformation filtering on each frame of speech signal according to the spectral line energy Carry out logarithmic transformation to each frame of speech signal after filtering, obtain the logarithmic energy spectrum of each frame of speech signal; Carry out discrete cosine transform to the logarithmic energy spectrum of each frame of speech signal, obtain the logarithmic energy spectrum of each frame of speech signal Hyperbolic cepstral coefficients. The invention specifies the effective utilization method of speech spectrum information through asymmetric hyperbolic sinusoidal frequency scale transformation, strengthens the channel difference between original speech and recording and playback attack speech, and can improve the performance of recording and playback attack detection.
Description
technical field [0001] The invention relates to the technical field of acoustic signal processing, in particular to an acoustic feature extraction method for recording playback attack detection in voiceprint recognition. Background technique [0002] Voiceprint recognition is an identification technology based on biometrics, which can identify the speaker's identity through the speaker's voice characteristics. Another widely used biometric identification technology. However, the security application of the voiceprint recognition system must solve the problem of spoofing attacks, including speech synthesis spoofing attacks and recording playback spoofing attacks. Due to the high similarity between the recorded and played back voice and the original voice, the biggest challenge is the recording and playback attack. [0003] Early recording and playback attack detection Due to the lack of public large corpus databases and baseline systems, it is difficult for developers to car...
Claims
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More Application Information
Patent Timeline
Login to View More Patent Type & Authority Patents(China)
IPC IPC(8): G10L17/02G10L17/22G10L19/02G10L19/26G10L25/24
CPCG10L17/02G10L17/22G10L19/02G10L19/26G10L25/24
Inventor 俞一彪郭星辰
Owner 上海企创信息科技有限公司