Phoneme-Aware Gain Control for Natural Speech and Lower Noise
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional gain control systems for audio communication and speech recognition fail to account for sound pressure differences between phonemes, leading to unnatural speech output and emphasizing noise, while existing methods for speech recognition are computationally expensive and inefficient.
Innovation Solution
A gain control system that uses speech models with feature and sound pressure information for each phoneme or cluster, calculating a weighted average of sound pressure ratios to adjust the gain and compensate for sound pressure, thereby reflecting phoneme characteristics and reducing noise emphasis.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If conventional gain control systems output constant sound pressure without considering phoneme differences, then the output sound pressure is uniform and easy to process, but the speech becomes unnatural and consonants are excessively emphasized
Solution Approach 1:
The patent applies local quality by assigning different sound pressure levels to different phoneme categories. Instead of uniform gain control, the system categorizes phonemes into groups (such as vowels, consonants, nasals) and applies specific sound pressure adjustments to each category, preserving the natural characteristics of each phoneme type while maintaining overall control.
Solution Approach 2:
The system changes the sound pressure parameter dynamically based on phoneme identification. By detecting the phoneme type and adjusting the sound pressure level accordingly, the system transforms the static constant sound pressure approach into a dynamic parameter adjustment strategy that adapts to the specific phoneme being processed.
2Productivity
If conventional gain control systems apply constant sound pressure to all phonemes, then the processing is simple and computationally efficient, but noise in the frequency band is excessively emphasized
Solution Approach 1:
The patent applies local quality by differentiating treatment between speech components and noise based on their frequency band characteristics. The system identifies the frequency band of the target speech and applies gain control selectively to this band, while applying different or reduced gain to noise frequency bands, thereby suppressing noise emphasis while maintaining processing efficiency.
3Manufacturing precision
If multiple speech models are used for different sound pressures to achieve accurate phoneme-specific control, then speech naturalness is improved, but the device complexity and computational cost increase significantly
Solution Approach 1:
The patent applies universality by using a single speech model that can adapt to different phoneme types through phoneme identification and classification. Instead of maintaining separate speech models for each phoneme or sound pressure level, the system uses one versatile model that processes different phoneme categories by applying appropriate sound pressure adjustments based on the identified phoneme type, thereby reducing the number of models required while maintaining accuracy.
Data Source
AI summary
Disclosed is a gain control system in which speech model constituted from a sound pressure and a feature is stored in a speech model storage unit for each of a plurality of phonemes or for each of clusters into which a speech is divided. When an input signal is given, a feature conversion unit calculates a feature and a sound pressure of the input signal. A sound pressure comparison unit determines a sound pressure ratio between the input signal and each of speech models. A distance calculation unit calculates a distance between the feature of the input signal and the feature of each of the speech models. A gain calculation unit calculates a gain value from the sound pressure ratio and information on the distance. A sound pressure compensation unit thereby compensates for the sound pressure of the input signal.