Bird species identification method based on dual-channel neural network

Abstract

The invention discloses a bird species identification method based on a dual-channel neural network. The bird species identification method comprises the steps that bird tweet signals of the known species are acquired and the preprocessed bird tweet signals are obtained through filtering, pre-emphasis and segmentation processing; signal sonagrams are generated based on linear frequency modulation wavelet transform; the bird tweet signals meeting the preset frame length range are cut to act as bird tweet time domain signals; the constructed preliminary identification model is trained with the signal sonograms acting as the input signals of a first channel, the bird tweet time domain signals acting as the input signals of a second channel and the corresponding bird species of the bird tweet signals acting as the identification result so as to obtain a bird species identification model; and the signals obtained by performing the same processing on the bird tweet signals to be identified are substituted in the bird species identification model to be identified so as to obtain the identification result. According to the bird species identification method based on the dual-channel neural network, the time domain characteristics and the time frequency characteristics of the tweet signals are fully utilized so that the efficiency and the accuracy of bird species identification can be enhanced.

Description

technical field [0001] The invention relates to the technical field of species recognition, in particular to a method for bird species recognition based on a dual-channel neural network. Background technique [0002] Birds are important indicators for biodiversity monitoring and ecological environment impact assessment. Through the investigation and monitoring of bird species, we can understand the status quo of bird resources, summarize the characteristics of bird species composition, quantity and diversity, and use these characteristics to directly reflect the environmental quality of habitats, the health of ecosystems, and biodiversity. conditions, the degree of disturbance of human activities to the ecosystem, and the degree of impact of land use and landscape changes on the ecosystem. Traditional bird survey methods are mainly manual survey methods, including line transect method, sample point method and direct counting method. Such methods require a lot of manpower, ...

AI Technical Summary

Problems solved by technology

[0004] Currently commonly used bird song classification and recognition methods include: 1. Classification methods based on template matching, the most representative of which is the dynamic time warping algorithm. Although this method has high recognition accuracy, the computational complexity is too large, which affects the recognition efficiency.

2. Establish a feature-based classification model to achieve classification. Commonly used models or methods include hidden Markov models, Gaussian mixture models, support vector machines, random forests, autonomous neural networks, k-nearest neighbors, and integrated learning. Among these methods Manual extraction of suitable difference features is still a major bottleneck

[0005] Therefore, in the process of realizing the present application, the inventor found that the prior art has at least the following technical defects: the current bird species identification method needs to consume a lot of manpower and material resources, and the accuracy of the identification results is not high, and it is difficult to obtain a comparative good recognition effect

Images