Method of recognizing large-sized vehicles on expressway based on GMM-HMM (gaussian mixture model and hidden Markov model)

Abstract

The invention provides a method of recognizing large-sized vehicles on an expressway based on GMM-HMM (gaussian mixture model and hidden Markov model) and relates to the technical field of intelligenttraffic. The method includes extracting MFCC (Mel frequency cepstral coefficient) features of an audio signal, analyzing a specific distribution of the features of the audio signal in a feature space, fitting a feature vector through Gaussian probability-density function to obtain an GMM-HMM, training the GMM in a training phase through EM (expectation maximization) algorithm to estimate model parameters, selecting a training sample via K-means algorithm, training GMM-HMM parameters through Baum-Welch algorithm in combination with an observation sequence probability distribution fitted via the Gaussian probability-density function in order to generate a training model, extracting MFCC feature parameters of audio data to be recognized in a recognition phase, and subjecting the MFCC featureparameters to feature probability matching with various models in a model base via Viterbi algorithm, so that the model having maximum matching probability is the recognition result. The method herein is suitable for accurately recognizing large-sized vehicles on the expressway.

Description

technical field [0001] The invention belongs to the technical field of intelligent transportation, in particular to a GMM-HMM-based identification method for large vehicles on expressways. Background technique [0002] Traffic information plays an important role in traffic management. Inadequate detection of expressway traffic information is an important reason for current expressway traffic congestion and frequent accidents. [0003] When a large vehicle is in a long-term transportation state on the highway, the safety performance of the large vehicle will be affected, such as problems in vehicle braking and operating the vehicle. It can cause tire deformation and puncture, brake failure, steering gear flicker and so on. Therefore, there are many potential safety hazards in the driving of large vehicles on expressways. Once a highway traffic accident occurs, it will seriously affect people's social and economic development and safe production, and the national economy wil...

AI Technical Summary

Problems solved by technology

There are many classification algorithms for sound. The commonly used classification algorithms include Nearest Neighbor (NN) and Dynamic Time Warping (DTW). Their main principle is to calculate the similarity between samples. According to the similarity The disadvantage of these algorithms is that the calculation and storage space needs to be large, because in the process of sound classification and recognition, the feature vectors of all training samples need to be stored

Images