Trajectory Analysis with Transportation Mode Classification

The patent introduces an intermediary processing layer that receives noisy location data from communications networks, applies filtering and validation rules, and produces cleaned location information. This intermediary process separates the raw data collection from the final analysis, allowing traffic monitoring to benefit from large datasets while protecting against measurement errors through multiple processing stages including outlier detection and data validation.

The patent performs preliminary cleaning, validation, and filtering of location data before it is used for traffic monitoring or analysis. By pre-processing the data to remove obvious errors, handle missing values, and validate data quality metrics beforehand, the system prepares the data in advance to reduce the impact of noise and inaccuracies on subsequent traffic monitoring operations.

The patent segments the mode of transport classification problem into distinct phases: data collection from multiple sources, feature extraction from trajectory patterns, model training with labeled data, and classification inference. By dividing the complex machine learning pipeline into separate modular components, the system achieves high classification accuracy while managing complexity through organized, reusable modules that can be independently optimized and maintained.

The patent utilizes parameter changes in trajectory data (such as speed, acceleration, direction changes, and temporal patterns) as input features for machine learning classification. By transforming raw location data into meaningful parameters that characterize different modes of transport, the system enables accurate classification while keeping the computational model relatively simple through feature engineering rather than requiring complex algorithms.

The patent merges multiple location data sources (cell tower triangulation, GPS, WiFi positioning, Bluetooth beacons) into a unified trajectory estimation framework. By combining data from these different sources with complementary strengths, the system achieves more accurate and reliable location estimates than any single source could provide alone, while managing the integration complexity through a standardized data model and fusion algorithms.

The patent creates a universal data processing framework that can handle multiple types of location data sources through a common interface and standardized processing pipeline. This multi-functional system can accommodate different data formats, accuracy levels, and update frequencies from various sources while applying consistent cleaning, validation, and fusion rules, thereby reducing the complexity that would otherwise arise from source-specific processing logic.