While todays interaction between man and machine is done many through screens a much more natural way to do so is via speech. Many industry leaders are already developing and improving algorithms that are meant to mimic and understand human language. Achieving human like performance will open a completely new way to interact with devices and offer new applications we haven’t thought of before. In this Project we test state of the art methods in machine learning using deep neural networks for “understanding” words and develop new architectures that try to capture all the features of human language.
We aim to build a Deep neural network that will be able to distinguish between 12 classes: 'Yes', 'No', 'Up', 'Down', 'Left', 'Right', 'On', 'Off', 'Stop', 'Go', Silence and Unknown. The input data to the net are 1 second recordings of different people saying a word. Below are schematics of the Networks tested.
Figure 1 Topology of the Basic network tested – raw data is transformed into spectrogram and then inserted into 8 convolutional layers each channel is then inserted into an LSTM and afterwards into a fully connected classifier
Figure 2 Two input convolutional network followed by an LSTM layer per channel
The Networks where tested on the Kaggle TensorFlow Speech Recognition Challenge – data set
