Tasks to do:
- Signals for the processing
- frame of voiced signal
frame of unvoiced signal
without header, fs=16000 Hz, for download to MATLAB
use function loadbin.m),
- longer utterances
(raw data without header, fs=16000 Hz),
- your signals from database zreratdb which is available for direct usage in CTU FEE computer rooms at the directory "G:\VYUKA\ZRERAT\signaly\zreratdb". For your work outside of CTU FEE, it is possible to download the archive zrerat_blocken_t01_cs0.zip which contains signals resampled to 16 kHz.
- on-line recorded signals (fs 16 kHz).
- LPC based encoding and generation (decoding) of voiced frame of the speech
- for voiced (unvoiced) speech frame compute parameters of autoregressive model of he order p=16, i.e. autoregressive coefficients a_k and power of the prediction error E_p.
- For voiced frame computed also value of pitch (basic period), i.e. f_0, T_0, and L_0 (period in samples).
- Create pulse or noise-based excitation for decoding of voiced or unvoiced speech frame.
- Create artificial voiced and unvoiced frame by filtering using prepared excitation and Synthesis filter of AR model.
- 1st checked result: for given frames of voiced and unvoiced speech frame_voiced.bin and
- time and frequency representation of the following signals: original, excitation, artificialy generated ones.
- LPC encoder for long signal
- Computed pitch for all short-time frames of longer signal including the detection of unvoiced and non-speech frames - speechpitch.m, med.m, see last seminar.
- For given signal, computed the parameters of AR model for all short-time frames
(autoregressive coefficients and power of the prediction error) - speechlpc.m, speechburg.m
- USe frame length 30 ms without overlapping (later frame length 30 ms and frame step 10 ms).
- 2nd checked result: matrix of encoded parameters of given signal SA106S06.CS0 - at each row of output matrix save f_0, E_p and vector of autoregressive coefficients a_k.
- LPC decoder for long signal
- Decode the signal from saved parameters of AR model. Be care and take into account the following problems:
- Filtering should be realized always without overlapping for the length related to the segmentation step of the encoding process.
- Do not forget to keep inicial conditions for the filtering of successive frames.
- Use artificcial excitation based on f0 saved for particular short-time frames.
Note. For the generation of excitation for successive frames kepp the basic periond accross the frames (i.e. the first pulse cannot be placed always at the first sample of short-time frame).
- 3rd checked result: decoded signal from previously computed matrix of encoded parametres for the utterance SA106S06.CS0 - observe waveform and spectrogram of original and decoded signals.
- encode and decode your own on-line recorded utterance
- 4th checked result: encoded and decoded on-line recorded signal - observe waveform and spectrogram of original and decoded signals.
- Try to use noise excitation only (i.e f0 = 0 for all short-time frames)
- Try to change (i.e. scale) f0 in voiced frames (f0 = scale*f0).