Intelligent Sound Processing for Selective Noise Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing noise reduction headphones in prior art technologies filter out both desired and undesired sounds, leading to a lack of selective noise reduction, which can be dangerous and inconvenient.
Innovation Solution
A sound processing apparatus for personal sound devices using machine learning-based models to classify scenarios, identify desired and undesired sounds, and filter audio signals accordingly, allowing only desired sounds to be heard while reducing ambient noise.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If passive noise reduction or active noise reduction is used to filter out noise, then noise reduction effect is improved, but desired sounds are also filtered out causing safety issues
Solution Approach 1:
The patent implements dynamic noise reduction by using machine learning models to continuously analyze audio signals and adjust filtering in real-time. The system dynamically identifies desired sounds (e.g., baby crying, station broadcasts) and adapts the noise reduction level accordingly, rather than applying static filtering. This resolves the contradiction by making the noise reduction effect variable - strong when only noise is present, and reduced when desired sounds are detected.
Solution Approach 2:
The system changes the parameter of noise reduction intensity based on the type of sound detected. By using cepstrum analysis, voiceprint recognition, and keyword detection to identify different sound parameters, the system adjusts the filtering strength dynamically. For example, when a baby's cry is detected, the noise reduction parameter is reduced to let the desired sound through, while maintaining strong noise reduction for ordinary ambient noise.
2Adaptability or versatility
If machine learning-based sound processing is implemented to selectively filter sounds, then sound processing intelligence is improved, but device complexity increases
Solution Approach 1:
The patent segments the sound processing function into multiple specialized machine learning models: a first model for scenario classification, a second model for sound identification (using cepstrum analysis, voiceprint recognition, keyword detection), and a third model for filtering control. This segmentation allows each model to focus on a specific task, improving overall intelligence while managing complexity through modular architecture.
Solution Approach 2:
The patent introduces an intermediary processing layer that bridges the gap between raw audio input and final noise reduction output. The intermediary consists of the machine learning models that analyze audio signals and generate control parameters for the noise reduction algorithm. This intermediary layer enables intelligent sound processing without directly modifying the core noise reduction hardware, thus managing device complexity.
Data Source
AI summary
A sound processing apparatus includes a receiving module configured to receive audio signals of one or more sounds acquired by a personal sound device, a processing module configured to use a sound processing model to perform: classification processing in which a type of a scenario where a user of the personal sound device is located is determined based on the audio signals; identification processing in which each of the one or more sounds is determined as a desired sound or an undesired sound based on the determined type of the scenario, and filtering processing in which filtering configuration is performed based on a result of the identification processing. The audio signals are filtered based on the filtering configuration, and an output module is configured to output the filtered audio signals, so as to provide same to the user.