Voice Command Authentication Using SNR and MFCC Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current voice command authentication methods face challenges in accurately distinguishing between legal and illegal users, particularly in noisy environments, due to limitations in signal processing and noise resistance.
Innovation Solution
A computer-implement voice command authentication method that calculates a Signal-to-Noise Ratio (SNR) value, converts sound signals into Mel-Frequency Cepstral Coefficients (MFCC) streams, and uses Dynamic Time Warping (DTW) distance and Log-likelihood ratio (LLR) values based on Gaussian Mixture Model with Universal Background Model (GMM-UBM) scores to determine whether a sound signal stream is from a legal or illegal user.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional voice authentication methods are used, then the authentication process is simple, but the accuracy is low especially in noisy environments
Solution Approach 1:
The authentication process is divided into multiple independent stages: SNR calculation for noise assessment, MFCC extraction for feature representation, DTW for temporal alignment, and GMM-UBM for probabilistic matching. Each stage processes specific aspects of the voice signal independently, improving overall accuracy while maintaining manageable complexity through modular design.
Solution Approach 2:
The patent transforms the voice authentication problem from direct signal comparison into a multi-dimensional feature space using MFCC coefficients. This dimensional transformation allows the system to capture spectral characteristics more effectively and perform authentication based on compressed spectral features rather than raw waveforms, significantly improving noise resistance.
2Reliability
If noise resistance is improved through advanced signal processing, then authentication accuracy increases, but processing time increases
Solution Approach 1:
The system pre-calculates and stores GMM-UBM model parameters during an enrollment phase before actual authentication is needed. This preliminary action prepares the reference models in advance, so that during real-time authentication, the system only needs to perform forward matching operations rather than training models, significantly reducing processing time while maintaining noise resistance.
Solution Approach 2:
The patent extracts only the essential spectral features using MFCC (Mel-Frequency Cepstral Coefficients) from the full voice signal, discarding redundant information. This extraction focuses computation on the most discriminative features for voice authentication, reducing processing time while preserving noise resistance by concentrating on spectral characteristics that remain stable in noisy environments.
3Measurement precision
If multiple authentication parameters are used, then false alarms are reduced, but the complexity of determination increases
Solution Approach 1:
The system employs feedback mechanisms where the SNR value influences the threshold selection for authentication decisions. When SNR is low (noisy environment), the system adjusts acceptance criteria accordingly, and when SNR is high, it can be more stringent. This feedback-based adaptive determination reduces false alarms while keeping the decision logic manageable through dynamic threshold adjustment rather than complex multi-parameter logic.
Data Source
AI summary
A computer-implement voice command authentication method is provided. The method includes obtaining a sound signal stream; calculating a Signal-to-Noise Ratio (SNR) value of the sound signal stream; converting the sound signal stream into a Mel-Frequency Cepstral Coefficients (MFCC) stream; calculating a Dynamic Time Warping (DTW) distance corresponding to the MFCC stream according to the MFCC stream and one of a plurality of sample streams generated by the Gaussian Mixture Model with Universal Background Model (GMM-UBM); calculating, according to the MFCC stream and the sample streams, a Log-likelihood ratio value corresponding to the MFCC stream as a GMM-UBM score; determining whether the sound signal stream passes a voice command authentication according to the GMM-UBM score, the DTW distance and the SNR value; in response to determining that the sound signal stream passes the voice command authentication, determining that the sound signal stream is a voice stream spoken from a legal user.