Ear-worn Audio Enhancement Using Overlapping Segment Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional neural network-based audio enhancement techniques for ear-worn devices, such as hearing aids, incur high latency due to sequential processing, making them unsuitable for in-person communication environments where low latency is desirable.
Innovation Solution
An ear-worn device processes audio signals by dividing them into overlapping segments, enhancing the overlapping portion immediately using a neural network engine (NNE), and then enhancing and outputting the non-overlapping portion during playback, thereby reducing latency by utilizing NNE outputs from previous segments to predict enhancements for subsequent segments.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If sequential processing is used for neural network-based audio enhancement, then processing accuracy is improved, but latency increases significantly
Solution Approach 1:
The audio signal is divided into overlapping segments that are processed in parallel. The neural network processes non-overlapping portions sequentially to maintain accuracy, while overlapping portions are enhanced using predictions from previous segments, enabling parallel processing that reduces latency without sacrificing enhancement quality.
Solution Approach 2:
The system performs preliminary enhancement of overlapping portions using predictions from previous segment processing before the complete audio signal is fully processed. This allows the enhanced audio to be output sooner, reducing latency while maintaining accuracy through the prediction mechanism.
2Productivity
If the audio signal is processed in real-time with neural network enhancement, then communication suitability is improved, but processing complexity increases
Solution Approach 1:
By segmenting the audio signal into overlapping portions and non-overlapping portions, the system enables real-time processing of critical overlapping sections while maintaining the ability to perform comprehensive enhancement on non-overlapping sections, thus achieving real-time capability without requiring the entire system to be equally complex.
Solution Approach 2:
The system applies full neural network processing to non-overlapping portions for maximum accuracy, while using prediction-based enhancement for overlapping portions. This partial application of complex processing only where needed enables real-time performance without the full complexity burden across the entire audio stream.
Data Source
AI summary
According to some embodiments, an ear-worn device, e.g., a hearing aid, is provided that operates to enhance audio signals detected by the ear-worn device. In some embodiments, the ear-worn device includes a microphone, a processing circuit coupled to the microphone, and an output signal generator coupled to the processing circuit. In some embodiments, a method for enhancing audio signals includes: detecting an audio signal with the microphone; as the audio signal is being detected, dividing the audio signal into a plurality of segments; enhancing the detected audio signal with the processing circuit of the hearing aid; and outputting the enhanced audio signal with the output signal generator. In some embodiments, enhancing the detected audio signal includes processing one or more of the segments of the detected audio signal with a neural network engine (NNE) of the processing circuit to obtain an output for enhancing the detected audio signal.