Ultrasonic Obstacle Detection With Neural Signal Compression
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current ultrasonic sensor systems for vehicle parking aids face challenges in transmitting data efficiently due to limited bandwidth, leading to potential data loss and misinformation, which can result in accidents. Existing methods do not effectively compress and decompress sensor signals to enhance data transmission without increasing the data rate.
Innovation Solution
A method utilizing artificial neural networks to extract feature vectors from sensor signals, which are then processed through multiple neural networks for obstacle detection and compression, allowing for the transmission of compressed data to a central computer system for fusion with other sensor data, reducing the need for increased data rates and enabling accurate obstacle detection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If sensor data transmission bandwidth is increased to improve obstacle detection accuracy, then measurement precision is improved, but device complexity and data transmission requirements increase
Solution Approach 1:
The patent extracts only the essential features from raw sensor data using neural networks before transmission. The sensor control unit performs feature extraction on ultrasonic sensor signals, identifying relevant obstacle information while discarding redundant data, thereby maintaining detection accuracy while reducing transmission bandwidth requirements
Solution Approach 2:
The patent performs preliminary data processing and feature extraction at the sensor level before transmission occurs. By pre-processing the data in the sensor control unit and transmitting only processed feature vectors rather than raw sensor signals, the system prepares data in advance to minimize transmission requirements
2Productivity
If data compression is applied to reduce transmission bandwidth, then data transmission efficiency is improved, but information loss may occur reducing detection accuracy
Solution Approach 1:
The patent transforms raw sensor data into feature vectors by changing the parameter representation. Neural networks process the original sensor signals and output compressed feature vectors that capture essential obstacle information in a different parameter space, achieving compression while preserving critical detection data
Solution Approach 2:
The system uses neural networks trained to recognize obstacle patterns, providing intelligent feedback on what information is essential to preserve during compression. The feedback mechanism ensures that only non-critical information is compressed or discarded, maintaining detection accuracy
3Quantity of substance
If feature extraction and neural network processing are performed at the sensor level, then data transmission requirements are reduced, but sensor system complexity increases
Solution Approach 1:
The patent divides the data processing function into segments: the sensor control unit performs feature extraction on individual sensor channels, then transmits only the extracted features to the central computer. This segmentation allows distributed processing that reduces overall transmission volume while managing complexity through functional division
Solution Approach 2:
The sensor control unit acts as an intermediary between the ultrasonic sensors and the central computer system. It performs neural network processing and feature extraction in this intermediate layer, transforming raw sensor data into compressed feature vectors before transmission, thereby reducing the data burden on the central system
Data Source
Figure 1
Figure 2
Figure 3~3(d)
AI summary
The invention relates to a method for operating an ultrasonic sensor, the associated sensor, the transmission of data to a computer system, and the associated decompression method in the computer system which controls the ultrasonic sensor. The compression method comprises the steps of detecting an ultrasonic received signal of an ultrasonic transducer, providing a signal-free ultrasonic echo signal model (610), carrying out at least once and optionally repeating a number of times the following steps of subtracting a reconstructed ultrasonic echo signal model (610) from the ultrasonic received signal (1) and forming a residual signal (660), the step of carrying out a method for recognising signal objects in the residual signal (660), supplementing the ultrasonic echo signal model (610) by the parameterised signal curve of a recognised signal object (600 to 605); and also ending the repetition of these steps if the values of the residual signal are below the values of a predetermined threshold signal. This is followed by transmitting at least some of the recognised signal objects in the form of symbols for these recognised signal objects to the computer system and using at least some of the recognised signal objects therein. These are preferably reconstructed in the computer system to give a reconstructed ultrasonic echo signal model.