Leveraging the urban soundscape: Auditory perception for smart vehicles

Abstract

Urban environments are characterised by the presence of distinctive audio signals which alert the drivers to events that require prompt action. The detection and interpretation of these signals would be highly beneficial for smart vehicle systems, as it would provide them with complementary information to navigate safely in the environment. In this paper, we present a framework that spots the presence of acoustic events, such as horns and sirens, using a two-stage approach. We first model the urban soundscape and use anomaly detection to identify the presence of an anomalous sound, and later determine the nature of this sound. As the audio samples are affected by copious non-stationary and unstructured noise, which can degrade classification performance, we propose a noise-removal technique to obtain a clean representation of the data we can use for classification and waveform reconstruction. The method is based on the idea of analysing the spectrograms of the incoming signals as images and applying spectrogram segmentation to isolate and extract the alerting signals from the background noise. We evaluate our framework on four hours of urban sounds collected driving around urban Oxford on different kinds of road and in different traffic conditions. When compared to traditional feature representations, such as Mel-frequency cepstrum coefficients, our framework shows an improvement of up to 31% in the classification rate.

Acoustic-Based Emergency Vehicle Detection Using Convolutional Neural Networks

“…The PBMs approach demonstrated better results than hidden Markov models (HMMs) trained on MFCC or log-mel features. Marchegiani and Posner [8] proposed a two-stage approach for acoustic-based EVD in smart vehicles, in which the first stage was to detect the presence of an abnormal sound, and the later stage involved noise reduction and classification. The framework in [8] borrowed the idea from image processing as it analyzed the spectrogram of the incoming signal as an image and employed spectrogram segmentation to isolate and extract the target signal from background noise.…”

From Bits of Data to Bits of Knowledge—An On-Board Classification Framework for Wearable Sensing Systems

“…Embedded machine learning is casually adopted in smartphones ( [6,7]), CCTV cameras (e.g., Reference [8]) and robots (e.g., Reference [9]). The concept of hierarchical classification itself has been proposed for sound classification in the context of smartphones [10] and smart vehicles [11]. In turn, intelligent duty cycling of high-power sensing elements, such as gyroscopes, GPS receivers and cameras, has been proposed as a means to extend the battery lifetime of smartphones [12,13] and mobile robots [14].…”

Speech Identification and Comprehension in the Urban Soundscape

“…Indeed, many major cities suffer from noise pollution [1], and there is significant evidence that this noise pollution contributes to several health disorders, including heart decease, diabetes and hearing loss, in millions of individuals [2,3]. At the same time, the urban soundscape is a rich source of information, and there is growing research interest in leveraging such information for realising the vision of autonomous driving in intelligent transportation systems [4]. Indeed, urban sound signals contain important cues that are vital for navigation in urban environments, yet difficult to obtain with traditional sensing modalities (e.g., cameras, lasers, etc.…”

“…Indeed, urban sound signals contain important cues that are vital for navigation in urban environments, yet difficult to obtain with traditional sensing modalities (e.g., cameras, lasers, etc. ), such as horns or sirens from emergency vehicles [4][5][6].…”