Audio Classification Feature Extraction for Natural Sound Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Consumer electronic devices face challenges in distinguishing between natural and artificial sounds, which is crucial for virtual assistants to accurately respond to human inputs and prevent false activations, especially in environments with multiple sound sources.
Innovation Solution
An electronic device equipped with microphones and a programmed processor that employs feature extractors to process audio signals, determining features such as directional information, sound classes, and distortion features, and uses a machine learning model to classify sounds as natural or artificial, with the option to access historical data for classification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If feature extractors process audio signals to determine multiple features (directional information, sound classes, distortion features), then the classification accuracy between natural and artificial sounds is improved, but the device complexity increases
Solution Approach 1:
The audio signal processing system is divided into multiple independent feature extractors, each responsible for extracting specific features (directional information, sound classes, distortion features). This segmentation allows parallel processing of different feature types, improving classification accuracy while managing computational complexity through modular architecture.
Solution Approach 2:
The system transforms the audio signal from a one-dimensional waveform into a multi-dimensional feature space by extracting multiple features simultaneously. This dimensional transformation enables the classifier to distinguish between natural and artificial sounds more effectively by considering multiple aspects of the audio signal rather than a single feature.
2Reliability
If a machine learning model classifies audio signals as natural or artificial, then the reliability of virtual assistant responses is improved, but the processing time increases
Solution Approach 1:
The feature extractors perform preliminary processing of the audio signal to extract relevant features before the machine learning classifier makes its decision. By pre-extracting directional information, sound classes, and distortion features, the system prepares the data in advance, reducing the processing time required during the actual classification step while maintaining high reliability.
Solution Approach 2:
The system incorporates historical data from previous classifications into the machine learning model, creating a feedback loop that continuously improves classification accuracy. This feedback mechanism allows the model to learn from past performance and refine its decisions, enhancing reliability over time while the feature extraction pipeline maintains efficient processing speeds.
3Speed
If the system processes audio signals in real-time to classify sounds, then the responsiveness of the virtual assistant is improved, but the computational resources required increase
Solution Approach 1:
The computational workload is segmented into multiple feature extractors that operate in parallel on different aspects of the audio signal. This segmentation allows the system to process audio in real-time by distributing computational tasks across multiple independent units, improving responsiveness while managing computational resource consumption through efficient parallel processing.
Solution Approach 2:
The system extracts only the necessary features (directional information, sound classes, distortion features) rather than processing every possible aspect of the audio signal. This partial action approach maintains real-time responsiveness by focusing computational resources on the most informative features for classification, reducing overall computational resource requirements while preserving system performance.
Data Source
AI summary
An electronic device has one or more microphones that pick up a sound. At least one feature extractor processes the audio signals from the microphones, that contain the picked up the sound, to determine several features for the sound. The electronic device also includes a classifier that has a machine learning model which is configured to determine a sound classification, such as artificial versus natural for the sound, based upon at least one of the determined features. Other aspects are also described and claimed.