Mel-spectrogram features for acoustic vehicle detection and speed estimation

Abstract

The paper addresses acoustic vehicle detection and speed estimation from single sensor measurements. We predict the vehicle's pass-by instant by minimizing clipped vehicle-to-microphone distance, which is predicted from the mel-spectrogram of input audio, in a supervised learning approach. In addition, mel-spectrogram-based features are used directly for vehicle speed estimation, without introducing any intermediate features. The results show that the proposed features can be used for accurate vehicle detection and speed estimation, with an average error of 7.87 km/h. If we formulate speed estimation as a classification problem, with a 10 km/h discretization interval, the proposed method attains the average accuracy of 48.7% for correct class prediction and 91.0% when an offset of one class is allowed. The proposed method is evaluated on a dataset of 304 urban-environment on-field recordings of ten different vehicles.

Video-Audio Neural Network Ensemble For Comprehensive Screening Of Autism Spectrum Disorder in Young Children

“…The resulting dataset consisted of 22 The neural network architecture is inspired by the 1-dimensional Convolutional Neural Network (1D-CNN) 74 with different audio features extracted from audio clips added to enhance classification 75 . These features include mean Mel Frequency Cepstral Coefficients (MFCCs) 41 , Mean Chromagram 44 , Mean Mel Spectrogram 40 , Mean Spectral Contrast 43 and Mean Tonal Centroid 42 . The features are concatenated into a one-dimensional vector by taking a mean along the time axis for each of the 10-second segments.…”

“…We then explored the potential of accurately predicting autism from the full audio band of the ADOS assessment. After normalizing the audio recordings and splitting them into 10-second segments, we extracted several acoustic features [40][41][42][43][44] that were then passed through a convolutional neural network (see Figure 2 and Methods). Following standard model hyperparameter tuning procedure, we deployed the 80-20 training validation split.…”

An approach to improving sound-based vehicle speed estimation

“…MA is predicted in a supervised fashion from the logmel spectrogram (LMS) of input audio. LMS represents a very popular feature in acoustic classification applications [32] and it proved very reliable in vehicle detection and speed estimation [3], [4], [8].…”

“…Large volumes of traffic data enable significant improvements in the performance of transportation, traffic safety and automatic traffic monitoring (TM) [1]. The TM data are used for valuable information extraction [2], which may include vehicle count [3], [4], shape [5], [6], speed [7], [8], acceleration [9], type [10], [11], plate number [12] and may be used to predict road accidents [13], [14].…”