pitch.py 文件源码

def process(self, spectrogram):
        mult = np.multiply(self.weights, np.ones((1, spectrogram.shape[1])))
        square_mag = np.power(spectrogram.magnitude(),2)
        square_mag = np.vstack((square_mag[1:,:],np.flipud(square_mag[1:,:])))
        square_mag_sum = np.sum(square_mag, 0)
        energy_threshold = np.percentile(square_mag_sum, 25)
        res = np.fft.rfft(square_mag.T, 2 * (self.n_bins - 1)).T
        resN = np.abs(res)
        resP = np.angle(res)
        yin = np.zeros((spectrogram.shape))
        yin[0,:] = 1
        tmp = np.zeros((spectrogram.shape[1],))
        for tau in range(1, self.n_bins):
            yin[tau,:] = square_mag_sum - resN[tau,:] * np.cos(resP[tau,:])
            tmp += yin[tau,:]
            yin[tau,:] *= tau/tmp

        tau = self.tau_min + np.argmin(yin[self.tau_min:self.tau_max,:],0)
        y_min = np.min(yin[self.tau_min:self.tau_max,:],0)
        tau = tau[:,np.newaxis]
        if self.interp:
            ranges = np.hstack((tau - 5, tau - 4, tau - 3, tau-2,tau-1,tau, 
                tau+1,tau + 2, tau + 3, tau + 4, tau + 5))
            new_tau = np.empty_like(tau)
            for frame in range(spectrogram.shape[1]):                    
                r = ranges[frame]
                r[0] = max(r[0], 0)
                r[1] = min(r[1], self.n_bins-1)
                val = yin[r.astype(int), frame]
                tck = interpolate.splrep(r, val, s=0, k = 2)
                xnew = np.arange(r[0],r[-1], (r[-1] - r[0]) /10)
                ynew = interpolate.splev(xnew, tck, der=0)
                new_tau[frame] = xnew[np.argmin(ynew)]
                y_min[frame] = np.min(ynew)            
            tau = new_tau
        tau = tau[:,0]
        pitch_confidence = 1- y_min
        pitch = np.zeros((spectrogram.shape[1],))
        pitch[tau!=0] = np.nan_to_num(self.sample_rate * 1.0 / (tau[tau!=0]))
        pitch[tau==0] = 0
        pitch_confidence[tau==0] = 0

        pitch[square_mag_sum < energy_threshold] = 0
        pitch_confidence[square_mag_sum < energy_threshold] = 0
        return (pitch, pitch_confidence)