The invention discloses a
shrimp call identification method based on periodic analysis and non-negative matrix factorization, and belongs to the technical field of cultivation. According to the method, a periodic analysis non-negative matrix factorization
algorithm with a self-learning capability is introduced, original acoustic data is decomposed into a basis matrix and a coding matrix, initial information about time change of each sound source is extracted,
natural environment noise and biological sound can be effectively separated, and the method has a strong self-adaptive capability to different signals and can be applied to the field of acoustic communication. According to the method, unsupervised
noise separation is realized, the accuracy and timeliness of recognition are improved, the method adapts to dynamic
environmental noise, and powerful data support is provided for
shrimp pond culture management; the method has a self-learning capability, can cluster different basis matrixes by introducing a
sparse constraint non-negative matrix factorization technology, effectively identifies and distinguishes the
sound production characteristics of shrimps, can meet
shrimp pond culture requirements, does not need to
train a model again when facing various biological
sound separation in various environments, and is efficient, rapid, high in accuracy and wide in application scene.