Video search method based on voice features
A video search and voice feature technology, applied in the field of video search based on voice features, can solve the problems of low voice search accuracy, unpredictable voice information, and difficulty in adding sufficient search algorithm accuracy, to optimize user experience, improve Applicability, the effect of improving accuracy
Active Publication Date: 2020-10-09
SICHUAN CHANGHONG ELECTRIC CO LTD
View PDF14 Cites 0 Cited by
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
[0004] The purpose of the present invention is to provide a video search method based on voice features, which is used to solve the unpredictable voice information in the prior art, and it is difficult to label enough data to ensure the accuracy of the search algorithm, resulting in inaccurate voice search. high technical issues
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 1
[0026] combine Figure 1 ~ Figure 3 Shown, a kind of video search method based on speech feature, comprises the following steps:
[0027] Step 1) Cut the voice information to be searched into multiple small segments, extract each segment of voice information into a feature vector, and each feature vector is a multidimensional feature vector, such as image 3 shown.
[0028] Step 2) Convert all extracted multi-dimensional feature vectors into text.
[0029] Step 3) The training set of the sequential neural network is set to the CA8 training set, and the samples of the CA8 training set are randomly cut according to the training set, verification set and training set: 8:1:1 ratio, and the reset gate of the sequential neural network is first initialized and update the parameters of the gate, and then import the text converted in step 2) into the sequential neural network. The sequential neural network uses the reset gate inside the gated neural unit to manage the data entering t...
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 a video search method based on voice features. The method comprises the following steps: extracting to-be-searched voice data into a plurality of multi-dimensional feature vectors; converting the extracted voice data of the multi-dimensional feature vector into a plurality of texts; importing the plurality of obtained texts into a sequential neural network for training to obtain a high-latitude feature vector; the convolutional neural network performs regression operation by taking the high-latitude feature vector as input; comparing the feature vector obtained by the convolutional neural network with a feature vector corresponding to a database, selecting a result with the highest similarity, and testing to obtain a final accuracy rate; and taking the video or audio as the output of the selected result with the highest similarity, completing the final search content, and feeding back the final search content to the user in the form of voice. According to the method and the device, the technical problem of low voice search accuracy caused by difficulty in marking labels with enough data to guarantee the precision of a search algorithm due to voice information change in the prior art is solved.
Description
technical field [0001] The invention relates to the field of artificial intelligence computer vision processing, in particular to a video search method based on voice features. Background technique [0002] The rapid development of the Internet has promoted the traditional real economy to become more and more intelligent, and people have begun to generate more demands that are close to life. As one of the mainstream communication applications of human-computer interaction in smart terminals, voice interaction is used more and more frequently in reality, and at the same time, the frequency of using smart terminals to search for multimedia resources is getting higher and higher. Existing smart terminals first convert speech into text, and search by fuzzy matching between character strings and fields in the library file. If there are fields in the database, the corresponding semantics can be found, and if there are no fields, the corresponding semantics cannot be found. This m...
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 IPC IPC(8): G06F16/783G06N3/04G06N3/08
CPCG06F16/7834G06N3/049G06N3/08
Inventor 梁敏
Owner SICHUAN CHANGHONG ELECTRIC CO LTD