DNN Audio Processing for Hearing Aid Speech Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional hearing aids struggle to distinguish between human speech and background noises, leading to amplified background noise interfering with speech reception and recognition in hearing-impaired patients.
Innovation Solution
A deep neural network (DNN) based audio processing method that uses a trained speech extraction model to selectively amplify speech signals while attenuating background noise, utilizing a mixed audio training dataset and user-specific gain compensation coefficients to adaptively process audio inputs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional frequency compensation algorithms are used to amplify audio signals in specific frequency ranges, then hearing compensation for hearing-impaired patients is improved, but background noise is also amplified along with speech, worsening speech reception and recognition
Solution Approach 1:
The patent extracts and separates speech signals from background noise using a deep neural network-based speech extraction model. The model processes mixed audio signals to isolate speech components, allowing the hearing aid to amplify only the extracted speech while suppressing background noise, thereby resolving the contradiction between effective hearing compensation and noise interference
Solution Approach 2:
The patent applies different processing strategies to different audio components: speech signals receive gain compensation and amplification tailored to the user's audiogram, while background noise components are attenuated or suppressed. This selective local processing allows simultaneous improvement of speech reception and reduction of noise interference
2Measurement precision
If deep neural network-based speech extraction model is used to separate speech from noise, then speech recognition is improved, but device complexity increases
Solution Approach 1:
The patent performs speech extraction and noise separation in advance through a trained deep neural network model before the audio signal is amplified. By pre-processing the audio signal to separate speech from noise components, the system simplifies subsequent processing steps and reduces the complexity of real-time decision-making in the audio amplification stage
3Measurement precision
If user-specific gain compensation coefficients are applied to compensate for individual hearing loss, then hearing compensation accuracy is improved, but processing time and computational resources increase
Solution Approach 1:
The patent pre-calculates and stores gain compensation coefficients based on user-specific audiograms before audio processing. These pre-computed coefficients are then rapidly applied during real-time audio processing, maintaining high compensation accuracy while minimizing processing time and computational burden during actual use
Data Source
AI summary
A deep neural network based audio processing method is provided. The method includes: obtaining a deep neural network based speech extraction model; receiving an audio input object having a speech portion and a non-speech portion, wherein the audio input object includes one or more audio data frames each having a set of audio data samples sampled at a predetermined sampling interval and represented in time domain data format; obtaining a user audiogram and a set of user gain compensation coefficients associated with the user audiogram; and inputting the audio input object and the set of user gain compensation coefficients into the trained speech extraction model to obtain an audio output result represented in time domain data format outputted by the trained speech extraction model, wherein the non-speech portion of the audio input object is at least partially attenuated in or removed from the audio output result.