diff options
Diffstat (limited to 'gr-vocoder/lib/codec2/phase.c')
| -rw-r--r-- | gr-vocoder/lib/codec2/phase.c | 32 |
1 files changed, 16 insertions, 16 deletions
diff --git a/gr-vocoder/lib/codec2/phase.c b/gr-vocoder/lib/codec2/phase.c index 0e1a14a60..69cc6697f 100644 --- a/gr-vocoder/lib/codec2/phase.c +++ b/gr-vocoder/lib/codec2/phase.c @@ -1,11 +1,11 @@ /*---------------------------------------------------------------------------*\ - - FILE........: phase.c - AUTHOR......: David Rowe - DATE CREATED: 1/2/09 - + + FILE........: phase.c + AUTHOR......: David Rowe + DATE CREATED: 1/2/09 + Functions for modelling and synthesising phase. - + \*---------------------------------------------------------------------------*/ /* @@ -22,7 +22,7 @@ License for more details. You should have received a copy of the GNU Lesser General Public License - along with this program; if not,see <http://www.gnu.org/licenses/>. + along with this program; if not,see <http://www.gnu.org/licenses/>. */ #include "defines.h" @@ -101,14 +101,14 @@ void aks_to_H( phase_synth_zero_order() - Synthesises phases based on SNR and a rule based approach. No phase + Synthesises phases based on SNR and a rule based approach. No phase parameters are required apart from the SNR (which can be reduced to a 1 bit V/UV decision per frame). The phase of each harmonic is modelled as the phase of a LPC synthesis filter excited by an impulse. Unlike the first order model the position of the impulse is not transmitted, so we create - an excitation pulse train using a rule based approach. + an excitation pulse train using a rule based approach. Consider a pulse train with a pulse starting time n=0, with pulses repeated at a rate of Wo, the fundamental frequency. A pulse train @@ -149,10 +149,10 @@ void aks_to_H( This E[m] then gets passed through the LPC synthesis filter to determine the final harmonic phase. - + Comparing to speech synthesised using original phases: - - Through headphones speech synthesised with this model is not as + - Through headphones speech synthesised with this model is not as good. Through a loudspeaker it is very close to original phases. - If there are voicing errors, the speech can sound clicky or @@ -207,21 +207,21 @@ void phase_synth_zero_order( G = 1.0; aks_to_H(model, aks, G, H, order); - /* + /* Update excitation fundamental phase track, this sets the position of each pitch pulse during voiced speech. After much experiment I found that using just this frame's Wo improved quality for UV sounds compared to interpolating two frames Wo like this: - + ex_phase[0] += (*prev_Wo+mode->Wo)*N/2; */ - + ex_phase[0] += (model->Wo)*N; ex_phase[0] -= TWO_PI*floor(ex_phase[0]/TWO_PI + 0.5); r = TWO_PI/GLOTTAL_FFT_SIZE; for(m=1; m<=model->L; m++) { - + /* generate excitation */ if (model->voiced) { @@ -254,7 +254,7 @@ void phase_synth_zero_order( A_[m].imag = H[m].imag*Ex[m].real + H[m].real*Ex[m].imag; /* modify sinusoidal phase */ - + new_phi = atan2(A_[m].imag, A_[m].real+1E-12); model->phi[m] = new_phi; } |