This is a project for Pattern Recognition course in Data Science and Machine Learning postgraduate programme in NTUA. Our goal is the implementation of a speech recognition system, that recognizes isolated words. The first part involves the extraction of the appropriate acoustic features from our recordings and their further analysis. These features are the cepstral coefficients, that are computed using a filterbank (inspired by psychoacoustic methods). More specifically, the system will recognize isolated digits in English. Our dataset contains dictations of 9 digits from 15 different speakers in separate .wav files. In total, there are 133 files, since 2 dictations are missing. The name of each file (e.g. eight8.wav) declares both the dictated digit (e.g. eight) and the speaker (speakers are numbered from 1 to 15). The sampling rate is Fs=16k and the duration of each dictation differs.
Nanos1/Speech-Recognition-using-HMMs-and-RNNs
Folders and files
| Name | Name | Last commit date | ||
|---|---|---|---|---|