Speech Processing Apparatus Using Weighted Air and Wearable Microphones
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Air microphones are vulnerable to noise and reverberation noise due to distance from the speaker, and wearable microphones perform poorly in speaker recognition applications due to high-frequency attenuation through skin or bone, leading to degraded speech signal quality.
Innovation Solution
A speech processing apparatus that combines recognition results from air microphones and wearable microphones, using weight calculations based on environmental conditions such as signal-to-noise ratio and distance between the microphone and speaker to improve recognition performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If an air microphone is used to acquire speech through air, then the speech signal can be obtained with full-band frequency, but the system becomes vulnerable to environmental noise and reverberation noise when the distance from the speaker is far
Solution Approach 1:
The patent combines air conduction microphones and bone conduction microphones into a unified speech processing system. The air conduction microphone captures full-band speech signals while the bone conduction microphone provides noise-robust speech signals, particularly effective when the speaker is far away. The system integrates both microphone types to leverage their complementary strengths and overcome their individual limitations.
Solution Approach 2:
The patent creates a composite microphone system that integrates two different conduction methods (air conduction and bone conduction) into a single speech processing apparatus. This composite approach allows the system to utilize both airborne sound waves and vibrations transmitted through bone, creating a more robust speech acquisition system that performs well across varying distances and noise conditions.
2Reliability
If a wearable microphone is used to acquire speech close to the speaker, then the system becomes less sensitive to surrounding noise, but the quality of speech signals is degraded due to high-frequency attenuation through skin or bone
Solution Approach 1:
The patent combines air conduction microphones and bone conduction microphones into a unified speech processing system. The air conduction microphone captures full-band speech signals while the bone conduction microphone provides noise-robust speech signals, particularly effective when the speaker is far away. The system integrates both microphone types to leverage their complementary strengths and overcome their individual limitations.
Solution Approach 2:
The patent applies different processing parameters and weightings to signals from air conduction and bone conduction microphones based on environmental conditions. When the speaker is close and noise is low, the system prioritizes air conduction signals for full-band quality. When the speaker is far or noise is high, the system increases reliance on bone conduction signals. This dynamic parameter adjustment optimizes speech signal quality across varying operational conditions.
3Device complexity
If only an air microphone is used for speech processing, then the system is simple in structure, but the performance degrades in noisy conditions and far distance from the speaker
Solution Approach 1:
The patent combines air conduction microphones and bone conduction microphones into a unified speech processing system. The air conduction microphone captures full-band speech signals while the bone conduction microphone provides noise-robust speech signals, particularly effective when the speaker is far away. The system integrates both microphone types to leverage their complementary strengths and overcome their individual limitations.
Solution Approach 2:
The patent implements dynamic weight adjustment mechanisms that adaptively change the contribution of each microphone type based on real-time environmental conditions. The weight decision unit dynamically determines optimal weighting factors for air conduction and bone conduction signals according to measured noise levels and speaker distance. This dynamic adaptation allows the system to maintain high performance across varying conditions without requiring complex manual configuration.
4Adaptability or versatility
If speech signals from different microphones are simply added together, then the system can utilize multiple microphone types, but the recognition performance does not necessarily improve because the combined signal becomes quite different from original signals
Solution Approach 1:
The patent applies different processing parameters and weightings to signals from air conduction and bone conduction microphones based on environmental conditions. When the speaker is close and noise is low, the system prioritizes air conduction signals for full-band quality. When the speaker is far or noise is high, the system increases reliance on bone conduction signals. This dynamic parameter adjustment optimizes speech signal quality across varying operational conditions.
Solution Approach 2:
The patent implements a feedback mechanism where the system measures environmental conditions (noise level, speaker distance) and uses this information to adjust the weighting of different microphone signals. The weight decision unit receives feedback from environmental sensors and recognition results, then dynamically adjusts signal weights to optimize recognition accuracy. This closed-loop feedback ensures that the combined signal remains optimized for accurate recognition rather than merely being a simple sum of different signals.
Data Source
AI summary
The speech processing apparatus 100 includes an air microphone speech recognition unit 101 which recognizes speech from an air microphone 200 acquiring speech through air, a wearable microphone speech recognition unit 102 which recognizes speech from a wearable microphone 300, a sensing unit 103 which measures environmental conditions, a weight decision unit 104 which calculates the weights for recognition results of the air microphone speech recognition unit 101 and the wearable microphone speech recognition unit 102 on the basis of the environmental conditions, and a combination unit 105 which combines the recognition results outputted from the air microphone speech recognition unit 101 and the wearable microphone speech recognition unit 102, using the weights.