Speaker Recognition Using Reverberant Environment Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Voice recognition systems in vehicles face challenges in maintaining accurate speaker recognition in highly reverberant environments, leading to potential accidental vehicle control by passengers rather than the intended driver.
Innovation Solution
An apparatus and method that calculates a speaker recognition score by selecting a reverberant environment model closest to the input data from multiple learning data sets and assigns weights based on the environment, using a voice feature extraction unit, reverberant environment probability estimation, and deep-neural-network learning to improve recognition accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a voice recognition device is designed to recognize the voice of many unspecified people, then the device can be used by multiple users, but the reliability of vehicle control deteriorates because passengers may be mistakenly recognized as speakers
Solution Approach 1:
The patent segments the speaker recognition system into multiple specialized recognition models, each trained for specific reverberant environments. Instead of using a single general-purpose recognition system, the device divides the recognition task across multiple environment-specific models (e.g., models for different vehicle interior acoustics), allowing accurate identification of the driver's voice while distinguishing it from passenger voices in various acoustic conditions.
Solution Approach 2:
The patent applies local quality by tailoring the speaker recognition characteristics to match specific reverberant environments. Each recognition model is optimized with local acoustic characteristics of particular vehicle interiors, enabling the system to adapt its recognition criteria to the specific acoustic properties of the environment, thereby reliably identifying the driver versus passengers.
2Volume of moving object
If the vehicle interior space is small, then the vehicle is compact and efficient, but the reverberation effect increases which deteriorates voice recognition accuracy
Solution Approach 1:
The patent applies preliminary action by pre-training multiple speaker recognition models during the development phase, with each model specifically trained on voice data collected from vehicles with different interior volumes and reverberation characteristics. This preliminary training ensures that when the system is deployed, it already has specialized models ready to handle the reverberation effects of compact vehicle interiors, maintaining high recognition accuracy without requiring real-time adaptation.
Solution Approach 2:
The patent utilizes parameter changes by varying the acoustic environment parameters (reverberation time, room volume, background noise levels) during the training phase of each recognition model. Each model is trained with specific parameter ranges corresponding to different vehicle types, enabling the system to maintain accurate speaker recognition despite the reverberation effects inherent in compact vehicle spaces.
Data Source
AI summary
Disclosed herein are an apparatus and method for recognizing a voice speaker. The apparatus for recognizing a voice speaker includes a voice feature extraction unit configured to extract a feature vector from a voice signal inputted through a microphone; and a speaker recognition unit configured to calculate a speaker recognition score by selecting a reverberant environment from multiple reverberant environment learning data sets based on the feature vector extracted by the voice feature extraction unit and to recognize a speaker by assigning a weight depending on the selected reverberant environment to the speaker recognition score.