Vehicle Ultrasonic AI Training for Multi-Object Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing ultrasonic signal evaluation methods in vehicles face challenges in achieving high detection accuracy and distinguishing between multiple objects at the same distance, particularly in the context of increasing competition for cost-effective and efficient components.
Innovation Solution
A computer-implemented method for training an AI module using a modified training dataset that includes features such as time-resolved signals, time-frequency representations, and source maps from ultrasonic array data, enabling improved object classification and detection accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional ultrasonic signal evaluation methods are used, then the system is simple and cost-effective, but the detection accuracy and object classification performance are insufficient
Solution Approach 1:
The patent replaces conventional signal processing methods with an AI-based evaluation system. The AI module is trained with modified training datasets that include time-resolved signals, time-frequency representations, and source maps from ultrasonic array data, enabling accurate object classification without complex manual signal processing pipelines.
Solution Approach 2:
The AI module is pre-trained offline with extensively prepared training datasets containing various object types and acoustic environments. This preliminary training allows the system to make accurate real-time classifications during vehicle operation without requiring complex runtime processing.
2Loss of information
If conventional ultrasonic signal evaluation methods are used, then the system is simple, but the ability to distinguish between multiple objects at the same distance is insufficient
Solution Approach 1:
The patent transforms ultrasonic signal data into multiple dimensional representations including time-resolved signals, time-frequency representations (spectrograms), and source maps. These multi-dimensional features enable the AI module to distinguish between multiple objects at the same distance by analyzing differences in their acoustic signatures across different dimensions.
Solution Approach 2:
The training dataset is segmented into multiple feature types (time-resolved signals, time-frequency representations, source maps, object labels). Each segment provides different information about objects, and the AI module learns to integrate these segmented features for accurate object differentiation.
3Measurement precision
If more sophisticated signal processing is applied to improve detection accuracy, then object classification improves, but computational effort and resource usage increase
Solution Approach 1:
Complex computational tasks including dataset generation, feature extraction, and model training are performed offline in advance. During actual vehicle operation, the pre-trained AI module performs only inference, which is computationally efficient and suitable for real-time processing in vehicle environments with limited computational resources.
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
Enhances the performance of environment sensing around vehicles by improving the ability to classify objects and differentiate between multiple objects at the same distance, reducing computational effort and resource usage.
Implementation Method 1
evaluating ultrasonic signals... at least one data entry about a reflection of an ultrasonic signal in an airborne sound range
Implementation Method 2
the ultrasonic signal can be an ultrasonic array signal that was ascertained by means of beamforming
Data Source
AI summary
A computer-implemented method for training an artificial intelligence module for determining an object in an environment of a vehicle. The method includes: providing a measured value dataset on a data carrier, wherein the measured value dataset comprises at least one data entry about a reflection of an ultrasonic signal in an airborne sound range and at least one data entry about a class of an object; generating a modified training dataset based on the measured value dataset, wherein generating the modified training dataset comprises the following steps: creating an input dataset based on the data entry about the reflection of the ultrasonic signal in the airborne sound range, creating an output dataset based on the data entry about the class of the object, wherein the method further comprises training an AI module based on the modified training dataset.