Sound event detection method based on double-branch discriminant feature neural network

Abstract

The invention discloses a sound event detection method based on a double-branch discriminant feature neural network, and the method comprises the steps: carrying out the feature extraction of a data set containing a sound signal, obtaining a log-mel spectrogram data set, and dividing the log-mel spectrogram data set into a training set, a test set and a verification set; and a double-branch discriminant feature network model is established, and the double-branch discriminant feature network model comprises double-branch sampling, feature extraction, double-branch feature fusion and loss fusion: the test set and the verification set are used as the input of the trained model, and the output of the model is the sound event detection result of the data set. Comprising a sound event type contained in the audio and starting and ending time of the event. According to the invention, the discriminative features of the tail class and the difficult-to-distinguish class are obtained in a double-branch discriminative feature fusion mode, the class weight of the classifier is balanced to a certain extent, and the sound event detection effect is improved.

Description

technical field [0001] The invention belongs to the design and application of a neural network model, in particular to the application of a dual-branch discrimination feature neural network model. Background technique [0002] In recent years, with the development of network technology and the emergence of a large amount of audio data, people have found that sound event detection technology can bring great help to human life. Sound events are specific and useful information contained in the audio. For example, the whistle of a car contains the information that the car is approaching, the sound of the siren contains the information that there may be dangerous events around, and the sound of wind and rain contains the information of the weather environment. The identification of these information is very useful for human life. [0003] In the current research, researchers generally divide the sound event detection task into two sub-tasks: sound event classification and sound ...

AI Technical Summary

Problems solved by technology

Secondly, in real life, there will be multiple sound events in one audio, so the situation faced by sound event detection becomes very complicated, and the difficulty of recognition will be greatly increased.

And because there is no large-scale, complete and reliable data set for early sound event detection, the development of sound event detection has been greatly restricted.

[0004] With the emergence of AudioSet and its sub-datasets for sound event detection in the fields of autonomous driving, smart home and smart monitoring, people have gradually discovered that there are data imbalances in real-life sound datasets, and the similarity between data categories is large. Phenomena such as labels

However, the current research on sound event detection ignores these difficult classification phenomena caused by the data distribution and data characteristics of the data set.

These imaginings can increase the recognition difficulty of the model and produce misleading results, thereby reducing the accuracy of the classification task in sound event detection.

Images