Sound Source Separation Using Neural Network
Shreya Sose, Swapnil G. Mali, Shrinivas Padmakar Mahajan · 2019
We have given an unsupervised separation of the sound for combination of the unknown different sound in the channel based on the DNN i.e. deep neural network. here it can be separate the different voices with the assumption of the dissimilarity measures of the audio signal. It has been shown that the different sound can be classified and separate. Here uses DNN based algorithm to acquire a mapping of mixed signal to the recovered signal. Also, it gives a separation architecture where it constructs an DNN separation module from which the separation of sound is done. After the training and testing of the sound separation it is shows that the performance of the system is better than other sound separation system and its very useful in segregate the multiple mixed signal. sound separation has varied uses in most real-time applications. In this paper, we use the mask for training targets of DNN for speech separation. The experiments are done with adding different noise conditions. The evaluation of this done using STOI evaluation parameter.