Vehicle Audio-Based Environmental Modeling for Occluded Object Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing vehicle navigation systems rely heavily on visual information, which can be limited and obscured, leading to a need for a more comprehensive environmental model that incorporates non-visual data.
Innovation Solution
A method and system for creating an audio-based model of a vehicle's environment using sensed audio information, combined with non-audio data like visual, radar, and lidar signals, to detect objects and predict their location and movement, even when they are occluded or not visible.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If visual information is used for environmental modeling, then the model can be constructed using available sensing data, but the field of view is limited and partially obscured by other objects
Solution Approach 1:
The patent combines visual sensor data with audio sensor data to create a hybrid environmental model. Audio information from microphones is processed alongside camera images to detect and track objects, particularly those occluded from visual view. This merging of different sensing modalities compensates for the limited field of view by adding acoustic coverage of the surrounding environment.
Solution Approach 2:
The patent adds an acoustic dimension to the traditional visual environmental model. By incorporating audio data from multiple microphones, the system creates a multi-dimensional sensory model that includes both visual and acoustic information. This additional dimension allows detection of objects beyond the visual field of view, effectively expanding the environmental awareness.
2Reliability
If visual sensors are used for obstacle detection, then the system can identify objects within the field of view, but detection fails when objects are occluded or visibility is poor
Solution Approach 1:
The patent uses audio information as an intermediary to detect objects that are occluded from visual view. When visual sensors cannot detect an object due to occlusion, the audio processing system analyzes sounds from the environment to identify and track the occluded object. This intermediary acoustic data pathway enables continuous detection reliability regardless of visual conditions.
Solution Approach 2:
The patent creates a composite environmental model that integrates both visual and acoustic data streams. This composite approach combines the strengths of visual sensing (for clear, visible objects) with acoustic sensing (for occluded or audible objects), resulting in a more reliable detection system that functions under diverse conditions including occlusion and poor visibility.
3Loss of information
If only visual information is processed, then the system complexity remains manageable, but the environmental model is incomplete and limited
Solution Approach 1:
The patent implements a multi-functional sensing system where the vehicle's sensor suite serves multiple purposes. Visual sensors continue to provide primary environmental modeling, while audio sensors provide complementary detection capabilities. This universal approach allows the same system architecture to handle both visual and acoustic data processing, maximizing the utility of each sensor type without proportionally increasing complexity.
Solution Approach 2:
The patent applies partial audio processing rather than full audio analysis. The system processes audio data selectively to complement visual information, focusing computational resources on integrating acoustic data only where it provides additional environmental awareness. This partial action approach adds value to the environmental model while keeping the increase in system complexity manageable.
Data Source
AI summary
A method for providing an audio-based model of an environment of a vehicle, the method may include obtaining, during a driving session of a vehicle, sensed information about the environment of the vehicle; wherein the sensed information may include sensed audio information. The sensed information may also include at least one type of non-audio sensed information; and generating an audio-based model of the environment based, at least in part, on the sensed audio information.