Rapid audio searching method based on GPU (Graphic Processing Unit)
An audio and fast technology, applied in the field of retrieval, can solve problems such as slow retrieval speed
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
Method used
Image
Examples
specific Embodiment approach 1
[0019] Specific implementation mode one: the steps of this implementation mode are as follows:
[0020] Step 1: Initial: Determine whether there is feature information of the audio clip in the image processor GPU,
[0021] If not, proceed to step 2 to preprocess the audio stream data;
[0022] If yes, enter step 3, and perform vector sliding matching of feature information of audio clips;
[0023] Step 2: Preprocessing: The central processing unit CPU divides the audio stream data input into the audio retrieval system into audio segments, performs feature extraction on each audio segment, and groups the feature information of the audio segments, and then classifies the feature information of each group of audio segments The information is sequentially transferred to the texture memory of the image processor GPU;
[0024] Step 3: audio segment vector sliding matching: the vector sliding matching module in the texture memory of the image processor GPU utilizes the segment vect...
specific Embodiment approach 2
[0032] Embodiment 2: The difference between this embodiment and Embodiment 1 is that the feature information of the audio segment includes Mel cepstral coefficients and their differential features and segment vector features, wherein the segment vector features are Mel cepstral coefficients and their differential features Dimensionality reduction features; other steps are the same as in the first embodiment.
specific Embodiment approach 3
[0033] Specific embodiment three: the difference between this embodiment and specific embodiment one or two is that the Mel cepstral coefficient feature matrix matching module and the vector sliding matching module are obtained by the following method:
[0034] Step A: The central processing unit CPU establishes an original audio library according to the function and scale of the audio retrieval system; performs feature extraction on each audio file in the original audio library, thereby obtaining Mel cepstral coefficients and their differential features and segment vectors Features two kinds of feature information, using the feature information to establish a reference template library;
[0035] Among them, the calculation of the Mel cepstral coefficients and their differential feature information is to convert the time-domain signal into a frequency-domain signal by using Fast Fourier Transform (Fast Fourier Transform, referred to as FFT), and then the pair of the frequency-d...
PUM
Login to View More Abstract
Description
Claims
Application Information
Login to View More