文献类型：期刊文献

英文题名：An Efficient Time-Domain End-to-End Single-Channel Bird Sound Separation Network

作者：Zhang, Chengyun[1] Chen, Yonghuan[1] Hao, Zezhou[2] Gao, Xinghui[1]

第一作者：Zhang, Chengyun

通信作者：Gao, XH[1]

机构：[1]Guangzhou Univ, Sch Elect & Commun Engn, Guangzhou 510006, Peoples R China;[2]Chinese Acad Forestry, Res Inst Trop Forestry, Guangzhou 510520, Peoples R China

年份：2022

卷号：12

期号：22

外文期刊名：ANIMALS

收录：;Scopus(收录号：2-s2.0-85142459655);WOS:【SCI-EXPANDED(收录号:WOS:000887018100001)】；

基金：This work is supported by the National Natural Science Foundation of China (32171520) and the Research Project of the Education Bureau of Guangzhou (No.202032882).

语种：英文

外文关键词：bird sound separation; transformer; deep learning; lower computational resources; dual-path network

摘要：Simple Summary Automatic bird sound recognition using artificial intelligence technology has been widely used to identify bird species recently. However, the bird sounds recorded in the wild are usually mixed sounds, which can affect the accuracy of identification. In this paper, we utilized massive amounts of data of bird sounds and proposed an efficient time-domain single-channel bird sound separation network. Our proposed network achieved good separation performance and fast separation speed while greatly reducing the consumption of computational resources. Our work may help to discriminate individual birds and study the interaction between individual birds, as well as to realize the automatic identification of bird species in various mobile and edge computing devices. Bird sounds have obvious characteristics per species, and they are an important way for birds to communicate and transmit information. However, the recorded bird sounds in the field are usually mixed, which making it challenging to identify different bird species and to perform associated tasks. In this study, based on the supervised learning framework, we propose a bird sound separation network, a dual-path tiny transformer network, to directly perform end-to-end mixed species bird sound separation in the time-domain. This separation network is mainly composed of the dual-path network and the simplified transformer structure, which greatly reduces the computational resources required of the network. Experimental results show that our proposed separation network has good separation performance (SI-SNRi reaches 19.3 dB and SDRi reaches 20.1 dB), but compared with DPRNN and DPTNet, its parameters and floating point operations are greatly reduced, which means a higher separation efficiency and faster separation speed. The good separation performance and high separation efficiency indicate that our proposed separation network is valuable for distinguishing individual birds and studying the interaction between individual birds, as well as for realizing the automatic identification of bird species on a variety of mobile devices or edge computing devices.