Speech Feature Extraction Using Inception Convolution Structure
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current speech recognition technologies using Neural Network Models have a low accuracy rate due to the absence of a specialized feature-extracting layer in the acoustic model.
Innovation Solution
The method involves performing spectrum analysis on speech to obtain a spectrum program, which is then processed using an Inception convolution structure from image recognition algorithms to extract speech features, improving recognition accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a Neural Network Model is used for speech recognition, then the system can perform speech recognition, but the accuracy rate is poor
Solution Approach 1:
The patent segments the speech recognition model into distinct functional components: a feature extraction layer that processes spectrum programs through multiple convolutional layers, and a separate recognition layer. This segmentation allows specialized feature extraction before recognition, improving accuracy while maintaining manageable complexity through modular design.
Solution Approach 2:
The patent transforms the speech recognition approach by adding a spectral dimension to the feature extraction process. Instead of processing raw speech signals directly, the system converts speech into spectrum programs (spectrograms) and applies image processing techniques, effectively adding a frequency-time visualization dimension that enhances feature discrimination and recognition accuracy.
2Measurement precision
If traditional speech recognition methods are used, then the process is simple, but the accuracy rate is poor
Solution Approach 1:
The patent introduces spectrum programs (spectrograms) as an intermediary representation between the raw speech signal and the recognition model. This intermediary transforms speech into a visual-like spectral format that can be processed by image processing techniques, bridging the gap between audio signal processing and feature extraction, thereby improving accuracy while using existing image processing tools.
Solution Approach 2:
The patent applies image processing algorithms (originally designed for visual data) to speech recognition tasks by treating spectrum programs as images. This universal application of convolutional neural networks and image processing techniques to audio data enables the system to leverage well-established, optimized algorithms across different modalities, improving accuracy without requiring entirely new methods.
Data Source
AI summary
Embodiments of the present disclosure provide a method and a device for extracting a speech feature based on artificial intelligence. The method includes performing a spectrum analysis on a speech to be recognized, to obtain a spectrum program of the speech; and extracting features of the spectrum program by using an Inception convolution structure of an image recognition algorithm, to obtain the speech feature of the speech. In embodiments, by performing the spectrum analysis on the speech to be recognized, the consecutive speech to be recognized is converted into the spectrum diagram. As the Inception convolution structure is an effective image recognition manner being able to accurately recognize features of an image, the spectrum program is recognized with the Inception convolution structure to extract the relative accurate speech feature from the speech to be recognized. Thus, the accuracy rate of the speech recognition is improved.