Voice recognition method and apparatus, and air conditioner
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current far-field voice recognition technologies face challenges with low recognition rates, long response times, and poor noise reduction due to limitations in microphone arrays and deep learning methods, which struggle to accurately locate sound sources and suppress noise in complex environments.
Innovation Solution
The integration of microwave radar technology to locate sound sources and adjust microphone array states, combined with a Long Short-Term Memory (LSTM) deep learning algorithm for training a far-field voice recognition model, enhances the accuracy and efficiency of voice recognition by improving sound source localization and noise suppression.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If deep learning methods or microphone array methods are used to remove reverberation and noise from far-field voice data, then noise reduction is attempted, but far-field voice recognition rate remains low and response time is long
Solution Approach 1:
The patent replaces traditional mechanical acoustic signal processing methods (microphone array beam forming) with electromagnetic wave-based microwave radar technology for sound source localization. The radar system emits microwave signals and receives reflections to precisely determine sound source position, which then guides selective activation of microphones. This substitution enables accurate localization in reverberant environments where acoustic methods fail, directly improving voice recognition rate by ensuring clean signal capture from the correct direction.
Solution Approach 2:
The patent introduces microwave radar as an intermediary device between the sound source and the microphone array. The radar acts as a mediator that first locates the sound source and provides directional information, which then controls which microphones should be activated. This intermediary approach allows the system to intelligently select and activate only the microphones closest to the sound source, improving both recognition accuracy and response time by avoiding processing from all microphones simultaneously.
2Measurement precision
If microphone array method is used with beam forming technology to suppress noise, then direction of sound source can be estimated, but the number of microphones and distance between them are limited, reducing direction range
Solution Approach 1:
The patent substitutes the mechanical limitation of microphone array geometry with electromagnetic wave-based radar measurement. Instead of relying on multiple microphones spaced at specific distances to achieve wide direction coverage, the microwave radar uses electromagnetic wave reflection to accurately measure sound source position regardless of physical array constraints. This allows precise direction estimation with fewer microphones, reducing device complexity while maintaining or improving measurement precision.
Solution Approach 2:
The patent makes the microwave radar system perform multiple functions: it not only locates the sound source direction but also determines distance to the sound source, and provides this information to control microphone activation. This multi-functionality replaces what would otherwise require separate systems (localization + distance measurement + microphone control), reducing overall device complexity while enhancing sound source direction estimation accuracy through the radar's superior ranging capabilities.
3Ease of operation
If general methods are used to process voice data without sound source location information, then processing is simpler, but noise reduction effect is poor and response time is long
Solution Approach 1:
The patent performs preliminary sound source localization using microwave radar before voice data processing begins. The radar determines the sound source position and provides directional information in advance, which then guides the selection and activation of appropriate microphones. This preliminary action enables the system to focus processing resources on the most relevant microphones, improving noise reduction effect by capturing clean signals from the correct direction while maintaining ease of operation through automated radar-guided microphone selection.
Solution Approach 2:
The patent implements a feedback loop where microwave radar continuously monitors sound source position and provides real-time directional information to the microphone control system. Based on this feedback, the system dynamically adjusts which microphones are activated to optimally capture the sound source. This feedback mechanism improves noise reduction effect by ensuring the system always uses the best microphones for current conditions, while maintaining processing simplicity through automated feedback-driven control.
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
This approach significantly improves far-field voice recognition performance by accurately locating sound sources, enhancing noise reduction, and shortening response times, resulting in higher recognition rates and better user experience in complex environments.
Implementation Method 1
acquiring sound source data including a location parameter of the sound source by combining microwave radar technology
Implementation Method 2
acquiring sound source data including a location parameter of the sound source by combining microwave radar technology
Data Source
AI summary
Provided is a voice recognition method and a voice recognition apparatus, and an air conditioner. The method includes: acquiring first voice data; adjusting, according to the first voice data, a collection state of second voice data to obtain an adjusted collection state, and acquiring the second voice data based on the adjusted collection state; and performing far-field voice recognition on the second voice data using a preset far-field voice recognition model so as to obtain semantic information corresponding to the acquired second voice data. The application can solve the problem in which far-field voice recognition performance is poor when a deep learning method or a microphone array method is used to remove reverberation and noise from far-field voice data, thereby enhancing far-field voice recognition performance.