Sound event localization and detection based on deep learning

Acoustic source localization(ASL)and sound event detection(SED)are two widely pursued independent research fields.In recent years,in order to achieve a more complete spatial and temporal representation of sound field,sound event localization and detection(SELD)has become a very active research topic.This paper presents a deep learning-based multioverlapping sound event localization and detection algorithm in three-dimensional space.Log-Mel spectrum and generalized cross-correlation spectrum are joined together in channel dimension as input features.These features are classified and regressed in parallel after training by a neural network to obtain sound recognition and localization results respectively.The channel attention mechanism is also introduced in the network to selectively enhance the features containing essential information and suppress the useless features.Finally,a thourough comparison confirms the efficiency and effectiveness of the proposed SELD algorithm.Field experiments show that the proposed algorithm is robust to reverberation and environment and can achieve higher recognition and localization accuracy compared with the baseline method.

Preliminary Study on Acoustic Characteristics of Ips grandicollis Larvae

[Objectives]The paper was to analyze the acoustic characteristics of Ips grandicollis larvae by pest acoustic detection technology,and to provide some reference for quarantine identification of pest larvae at ports.[Methods]The trial was performed in a self-invented insect sound recording container with good sound insulation effect.I.grandicollis larvae were placed separately on split P.ponderosa logs to observe and record the activities of larvae.AED-2010 was used for signal acquisition and SP-1 probe was used to collect signals at a distance of 5 cm from the larvae.The activity signals of larvae were intercepted,and the signal de-noising was further processed by Goldwave software.Finally,the acoustic signals were carried out correlation analysis by using MATLAB toolbox.[Results]I.grandicollis larvae had more regular feeding acoustic signal characteristics than crawling acoustic signal characteristics,and the two types of acoustic signal characteristics were quite different.The duration of feeding acoustic signal pulse of I.grandicollis larvae was 50-85 ms,the frequency was less than 1 KHz,and the signal frequency was mainly distributed in the range of 510.00-620.20 Hz.There was no obvious law in time domain features of larval crawling acoustic signals and the signal frequency was low,which was below 0.6 KHz and was mainly distributed in the range of 258.00-530.00 Hz.[Conclusions]It is feasible to carry out quarantine identification of I.grandicollis larvae by using feeding acoustic signals.It is suggested to select the feeding acoustic signals with obvious characteristics in the sound monitoring and identification of pests.

Auditory detection of sound signals with complex time-frequency characteristics

The mechanism of the human auditory system in detecting sound signals with complex time frequency charcteristics in a white noise background was reviewed and discussed.The efficiency of such auditory detection was assessed by comparing it with that of parallel visual detection of the output of an analogous model displayed on the oscilloscope screen. The results suggest that the detection model of the human auditory system is quite similar to a tone correlator when the time frequency characteristics of the signal are known and to an energy detector when the signal is unknown. The relationship between the threshold signal to noise ratio and the signal duration is derived for different time frequency characteristics.