Audio-Based Map Object Prediction Using Machine Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The creation and updating of digital maps require substantial manual effort due to the ever-changing nature of road networks and map objects, limiting the inclusion of new and updated map features in a timely and comprehensive manner.
Innovation Solution
A method and apparatus that utilize audio data generated by a vehicle interacting with map objects, employing a trained machine learning model to predict map objects based on audio amplitude and frequency features, allowing for efficient identification and inclusion of new or changed map objects in digital maps.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual methods are used to identify and update map objects, then map accuracy can be maintained through human verification, but the process requires substantial manual effort and time, limiting the timeliness of map updates
Solution Approach 1:
The patent replaces manual mechanical identification processes with an automated acoustic detection system. Microphones capture audio signals of map objects (e.g., railroad crossings, potholes), and machine learning models automatically identify and classify these objects, substituting human manual verification with automated acoustic analysis while maintaining identification accuracy
Solution Approach 2:
The system enables map objects to be automatically identified and updated through acoustic signatures without requiring manual intervention. The machine learning model processes audio data independently to detect, classify, and update map object information, allowing the system to self-update map data in real-time as vehicles pass over map objects
2Loss of information
If comprehensive manual surveying is conducted to include all map objects, then map completeness improves, but the substantial effort required limits the ability to rapidly identify and include new or changed map objects
Solution Approach 1:
The system continuously captures audio data as vehicles traverse the road network, enabling ongoing detection and identification of map objects without interruption. This continuous acoustic monitoring ensures that new or changed map objects are detected and added to the map database in real-time, maintaining map completeness without requiring periodic manual surveys
Solution Approach 2:
The acoustic detection system serves multiple functions: it identifies various types of map objects (railroad crossings, potholes, road seams, speed bumps), classifies them by type, and updates map databases automatically. This universal approach using audio analysis replaces multiple specialized manual surveying methods, achieving comprehensive map coverage through a single automated system
Data Source
AI summary
A method, apparatus and computer program product are provided to predict a map object based at least in part on audio data. With respect to predicting a map object, audio data created by a vehicle while driving over a road surface is obtained. The audio data includes one or more audio amplitude features, one or more audio frequency features or a combination of audio amplitude features and audio frequency features. The audio data including the one or more audio amplitude features is provided to a machine learning model and the map object that created the audio data while the vehicle interacted therewith is predicted utilizing the machine learning model. A method, apparatus and computer program product are also provided for training the machine learning model to predict the map object based at least in part on audio data.