Audio Signal Modulation Bandpass Filtering for Hearing Aid Distortion Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional dynamic compression methods for audio signals in hearing aids often amplify speech components equally to background noise, leading to unsatisfactory signal quality and potential distortions.
Innovation Solution
A method involving digital recording of audio signals, spectral time signals are divided into frequency bands, and a spectro-temporal representation is created through modulation bandpass filtering, followed by conversion using a compression function to enhance speech intelligibility by adjusting signal levels within specific hearing thresholds.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional dynamic compression methods are used to compress audio signals in hearing aids, then the dynamic range is reduced to make quiet sounds audible, but speech components are amplified equally to background noise leading to distorted output signals
Solution Approach 1:
The audio signal is divided into multiple frequency bands using filter banks, and each band is processed independently through modulation spectrum analysis. This segmentation allows selective compression of different frequency regions while preserving speech characteristics and preventing uniform amplification of noise across all frequencies.
Solution Approach 2:
Different compression parameters and modulation transfer functions are applied to different frequency bands and modulation ranges. Critical speech frequency bands maintain higher fidelity with less compression, while non-critical bands undergo more aggressive compression. This local differentiation ensures speech quality is preserved while still achieving dynamic range reduction.
2Reliability
If uniform dynamic compression is applied across all frequency bands, then the overall dynamic range is reduced, but speech intelligibility deteriorates due to equal amplification of speech and noise
Solution Approach 1:
The spectrum is divided into multiple frequency bands with speech-critical regions identified and protected. By segmenting the processing, the system can apply different compression strategies to speech and non-speech regions, preserving intelligibility while compressing the overall dynamic range.
Solution Approach 2:
Compression parameters such as compression ratio, threshold, and modulation transfer functions are dynamically adjusted based on the detected speech content and frequency band. Speech-critical bands receive parameter settings that preserve intelligibility, while other bands use more aggressive compression parameters.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The invention relates to a method for the computer-aided processing of audio signals. In the method according to the invention, an audio signal in the form of a time signal of amplitude values is digitally recorded as an input signal (A). The spectral time signals (S1, S2,..., Sn) contained in the input signal (A) are determined in a plurality of frequency bands (FB1, FB2,...,FBn) having associated signal frequencies (SF1, SF2,...,SFn) in the form of center frequencies of the respective frequency bands (FB1, FB2,...,FBn), whereby a plurality of signal levels is obtained for each spectral time signal (S1, S2,..., Sn). From the spectral time signals (S1, S2,..., Sn), a spectro-temporal representation (STR) of the input signal (A) is determined, which is subjected to a spectral modulation bandpass filtering (SBF). By means of the spectral modulation bandpass filtering (SBF), from a plurality of non-overlapping spectral modulation frequency bands (MB1, MB2,..., MB5), the spectral variation of the signal components contained in each modulation frequency band (MB1, MB2,..., MB5) along the signal frequencies (S1, S2,..., Sn) is extracted, whereby one or more first modulation signal representations (MSR) are obtained for each modulation frequency band (MB1, MB2,..., MB5). The first modulation signal representations (MSR) are converted into second modulation signal representations (MSR), on the basis of which a modified spectro-temporal representation (STR) is determined. Finally, an output signal (A') that is modified in comparison with the input signal (A) is produced from the modified spectro-temporal representation (STR1).