Audio Source Localization via Correlogram Probability Maps
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing time-difference-of-arrival techniques are primarily unsuitable for localizing audio sources such as human speech and many other sounds, as they rely on distinctive characteristics of transient or pulse-like sounds, making it difficult to determine the location of audio sources in environments where the origination time of sounds cannot be determined.
Innovation Solution
A system using multiple microphones positioned at known locations to calculate correlograms, which indicate the probability of audio sources at various distances, creating a probability map to localize audio sources, and tracking their positions over time, even for multiple concurrent sources like human speakers.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If time-difference-of-arrival techniques are used to localize audio sources, then the location of transient or pulse-like sounds can be determined, but the technique becomes unsuitable for localizing continuous sounds such as human speech
Solution Approach 1:
The patent transforms the audio signals from the time domain to the frequency domain using Fourier transforms, allowing analysis of continuous sounds like speech by examining frequency components rather than relying on transient time-domain features. This parameter transformation enables the system to handle both transient and continuous sound types effectively.
Solution Approach 2:
The patent introduces correlograms as an intermediary computational tool that compares frequency-domain representations of audio signals from multiple microphones. The correlogram calculation serves as a mediator that bridges the gap between raw audio signals and source location determination, enabling versatile localization across different sound types.
2Measurement precision
If time-of-flight measurements are used to determine audio source locations, then location information can be obtained, but the method requires knowing the origination time of sounds which is not available in many situations
Solution Approach 1:
Instead of using time-of-flight which requires knowing when sounds were generated, the patent inverts the approach by using time-difference-of-arrival at multiple microphone locations. Rather than measuring absolute time from source to detector, the system measures relative time differences between detectors, eliminating the need for origination time information.
Solution Approach 2:
The patent replaces the time-of-flight measurement mechanism with a frequency-domain correlation mechanism. By transforming audio signals to frequency domain and calculating correlograms, the system substitutes direct time measurement with a computational correlation approach that works with continuous sounds and does not require origination time.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
Enables effective localization of audio sources, including human speech and other sounds, by generating probability maps that accurately indicate the likelihood of sound sources within an environment, allowing for continuous tracking of users and sound sources.
Implementation Method 1
Time-difference-of-arrival uses microphones at multiple locations to detect arriving audio. Assuming that a discrete event can be detected in the audio, the time of arrival of that event may be compared between different microphones to determine the likely location of the audio source relative to the microphones.
Implementation Method 2
Correlograms are then calculated for different pairs of the microphones based on the audio signals captured by those microphones. Each correlogram indicates the degree of correlation between the audio signals received by the two microphones over a range of timing shifts between the audio signals.
Data Source
AI summary
Techniques are described for determining locations of audio sources. Audio signals are captured from multiple locations. Pairs of the audio signals are analyzed to create correlograms, indicating correlation scores corresponding to different time offsets between the signals. Based on the correlograms, various locations are analyzed to determine probabilities of audio originating from those locations. The highest probabilities are found, indicating locations containing audio sources. In some situations, the audio sources may be reflective sources, and the locations of the reflective sources may be used to determine locations of objects or surfaces.