What makes audio event detection harder than classification?

Abstract

Abstract-There is a common observation that audio event classification is easier to deal with than detection. So far, this observation has been accepted as a fact and we lack of a careful analysis. In this paper, we reason the rationale behind this fact and, more importantly, leverage them to benefit the audio event detection task. We present an improved detection pipeline in which a verification step is appended to augment a detection system. This step employs a high-quality event classifier to postprocess the benign event hypotheses outputted by the detection system and reject false alarms. To demonstrate the effectiveness of the proposed pipeline, we implement and pair up different event detectors based on the most common detection schemes and various event classifiers, ranging from the standard bag-of-words model to the state-of-the-art bankof-regressors one. Experimental results on the ITC-Irst dataset show significant improvements to detection performance. More importantly, these improvements are consistent for all detectorclassifier combinations.

Review of anomalous sound event detection approaches

“…In surveillance SED architectures consist of additional background subtraction,object tracking & situational analysis processes in the pipeline [8]. Latest research of an improved pipeline suggest a verification step to reduce false positives after the SED process [3]. Figure 1 shows the extended SED pipeline.…”

“…The Sound event detection (SED) research field has been an active recently [1][2][3][4]. Autonomous audio surveillance has become more efficient as that artificial intelligence has stepped up the game [5].…”

“…Acoustic surveillance specifically in security application is still new to the world and requires research to allow better performances [4,7,8]. The detection on anomalous audio events effectively while avoiding false positives is crucial in security [3]. Two categories of sound recognition, non-speech the determination of the sound event source and speech recognition of verbal language [9][10].…”

Audio-Based System for Automatic Measurement of Jump Height in Sports Science

“…Both classification and detection is used in this work, the first study to distinguish take-off from landing and the second to precisely detect, in time, the occurrence of both events. Even though it has been accepted that audio event classification is easier to deal with than detection, recent works emphasize the importance of detection [18], which is of particular importance in this work. In fact, as discussed later, distinguishing take-off from landing will be straightforward, so more effort will be put on precise time detection.…”

Unifying Isolated and Overlapping Audio Event Detection with Multi-label Multi-task Convolutional Recurrent Neural Networks

“…The research was performed when H. Phan was at the University of Oxford and supported by the NIHR Oxford Biomedical Research Centre. Corresponding author: h.phan@kent.ac.uk reduction [22,7]. Particularly, the multitasking approach that jointly performs event detection and event boundary estimation [23,6,24] has demonstrated state-of-the-art performance on different benchmark datasets.…”