Speech Feature Recognition for Privacy Scoring and Data Encryption
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems fail to effectively monitor and analyze speech data from network voice or video calls to protect enterprise confidentiality, allowing rival companies to learn sensitive information.
Innovation Solution
A deep learning-based feature recognition system that denoises, characterizes, and analyzes speech data to identify and encrypt sensitive information, using a character recognition model and privacy scoring to manage and secure enterprise data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If speech data is monitored and analyzed in real-time to protect enterprise confidentiality, then security and privacy protection are improved, but system complexity and processing requirements increase
Solution Approach 1:
The system pre-processes speech data through denoising and feature extraction before analysis, preparing the data in advance for more efficient processing. The character recognition model is pre-trained with phonetic alphabet strings and tone marking information, enabling faster real-time analysis while maintaining security requirements.
Solution Approach 2:
The patent introduces an intermediary character recognition model that converts speech data into text form before privacy analysis. This intermediary step simplifies the subsequent privacy scoring and analysis processes, making the overall system more manageable despite the increased security requirements.
2Measurement precision
If deep learning models are used to accurately recognize and analyze speech data, then recognition precision is improved, but computational resources and processing time increase
Solution Approach 1:
The speech recognition process is segmented into distinct stages: denoising, feature extraction (phonetic alphabet strings and tone markings), character recognition, and privacy analysis. This segmentation allows each component to be optimized independently, reducing overall computational burden while maintaining high accuracy through specialized processing at each stage.
Solution Approach 2:
The system transforms speech data from audio domain to feature domain by extracting phonetic alphabet strings and tone marking information. This parameter transformation converts complex audio signals into structured textual features, reducing computational complexity for subsequent recognition and analysis while preserving recognition accuracy.
3Loss of information
If all speech data is processed and stored for analysis, then data completeness is improved, but data security risks and storage requirements increase
Solution Approach 1:
The system extracts only the essential features from speech data - specifically phonetic alphabet strings and tone marking information - rather than storing and processing the complete audio data. This extraction approach maintains data completeness for recognition purposes while significantly reducing security risks and storage requirements by eliminating the need to retain sensitive audio content.
Solution Approach 2:
The system creates a textual copy of speech data through character recognition rather than storing the original audio. This copy contains the essential information needed for privacy analysis while being less sensitive and more secure to store and process, effectively replacing the need to handle and store sensitive audio data.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
The system effectively identifies and secures sensitive information, preventing data intrusion and ensuring enterprise development by automatically recognizing speech data, translating it into text, and encrypting sensitive content.
Implementation Method 1
performing a short-time Fourier transform on the speech signal to convert the speech signal to a frequency domain
Implementation Method 2
in the frequency domain, a filter is used to remove higher frequency noise components and retain the main speech signal
Data Source
AI summary
A feature recognition method and system based on deep learning comprises retrieving speech data to be recognized from a database, and obtaining speech data to be recognized. The invention retrieves speech data to be recognized from a database, denoising the data, establishing a character recognition model, inputting denoised speech data to be recognized, automatically recognizing pronunciation of each character in the denoised speech data and translating it into text, obtaining text data and proofreading, obtaining text data after proofreading, analyzing privacy of the speech text data and scoring privacy, comparing privacy scoring with predetermined threshold, determining whether speech text data has an impact on development of an enterprise, and encrypting original speech data and storing it in database to retain evidence and to know who is the informer, which prevents the data from intrusion and destruction of outside, so that system can meet development needs of enterprise.