A Speaker Recognition Method Based on Siamese Network Model and KNN Algorithm

Abstract

The invention discloses a speaker recognition method based on a twin network model and a KNN algorithm. Step S1: use a microphone to collect voice information of the speaker as a data set to train the RNN network model; step S2: use the trained RNN network to construct a twin network The model is combined with the KNN algorithm to identify the speaker. Adopt the technical solution of the present invention, train the data set of the speaker in the database, determine that the input of each speech signal input into the Siamese network can output the characteristics of the speaker, and use the cosine distance to calculate the difference between the different output feature vectors. The distance between them and the use of the KNN algorithm to determine whether they belong to the same speaker, so that a small number of samples can also identify the speaker, and with the increase in the number of speakers, the network does not need to be retrained, which reduces the number of data samples for the neural network requirements, while effectively improving the real-time and accuracy of speaker recognition.

Description

technical field [0001] The invention belongs to the technical field of human-computer interaction, in particular to the technical field of speaker recognition, and specifically designs a speaker recognition method based on a twin network model and a KNN algorithm. Background technique [0002] In the field of human-computer interaction, with the rapid development of artificial intelligence, pattern recognition and other technologies, the interaction between humans and computers is getting closer and closer. The traditional contact interaction methods can no longer meet people's needs. The interactive way of communication habits has become a research hotspot in recent years. As one of the main channels of human-computer interaction, speaker recognition has gradually become an important research topic in the field of interaction. [0003] Existing methods for speaker recognition mainly include methods of speech feature extraction and template matching technology, methods of s...

AI Technical Summary

Problems solved by technology

The method based on template matching is to pre-train the recognized voiceprint samples and match the voiceprint to be recognized with it. This method is simple to operate, but the recognition accuracy is not high and a large number of data samples are required.

In the method based on the speech statistical model, the recognition task is defined as the probability of calculating the variable. This method has high recognition accuracy, but requires a large amount of data for verification.

In the method based on deep learning technology, the neural network is used to capture the hidden features of the speaker so that it can better represent the speaker. This method not only requires a large amount of data, but also requires the neural network to be updated every time the data set is updated. Retraining is not conducive to new information entry

Images