Speech Recognition Consonant Noise Separation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional speech recognition methods for consonant signals struggle to accurately distinguish between consonant and noise signals, particularly for hearing-impaired individuals, as they rely on energy and zero-cross rate ratios which do not meet actual demand for accuracy.
Innovation Solution
A speech recognition apparatus and method utilizing a band-pass filter unit to separate voice signals into consonant frequency bands, followed by a processing unit that calculates energy ratios and zero-cross rates to determine whether a signal is a consonant or noise signal, employing weighted averages and preset ratio ranges to improve accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional determination methods using energy and zero-cross rate ratios are used, then the system can distinguish between consonant and noise signals, but the recognition accuracy fails to meet actual demand
Solution Approach 1:
The patent divides the voice signal into multiple frequency bands using band-pass filters (e.g., 500-2000 Hz for first consonant frequency band, 2000-4000 Hz for second consonant frequency band). This segmentation allows independent energy analysis of different frequency components, enabling more precise identification of consonant signals which have characteristic energy distributions across bands, thereby improving recognition accuracy and reducing mistaken identification with noise signals.
2Measurement precision
If multiple energy ratio calculations and preset ratio ranges are used, then the recognition accuracy for consonant signals is enhanced, but the device complexity increases
Solution Approach 1:
The patent calculates multiple energy parameters including total energy, energy in first consonant frequency band, energy in second consonant frequency band, and their ratios. By changing and analyzing multiple parameters simultaneously (energy ratios across different bands), the system achieves enhanced recognition accuracy. The processing unit compares these parameters against preset ratio ranges to determine consonant signals, which improves precision while managing complexity through systematic parameter analysis.
Data Source
AI summary
A speech recognition apparatus and a speech recognition method are provided. In the invention, whether an original voice sampling signal corresponding to a target voice frame is a noise signal is determined according to a ratio of an energy of a first consonant frequency band signal to an energy of a second consonant frequency band signal, a ratio of an energy of the first consonant frequency band signal to an energy of the original voice sampling signal and a ratio of an energy of the second consonant frequency band signal to an energy of the original voice sampling signal.