Freely-controlled sound-phase acoustics-variable speed (I)

Relatively antique VOICE: LPC voice, phase voice

The LPC phonograph uses the all-pole model to represent the voiced voice and approximate the voiced voice. According to the least square error minimization principle, a set of parameters of this model are obtained, namely, the LPC coefficient, which quantifies the coefficients, the data volume can be greatly compressed. the time domain analysis uses a group of sample points in the past to predict the current sample points.

It can be derived from short-term Fourier transformation and inverse transformation. A signal x (n) accumulates signals in various frequencies after filtering under certain conditions through a filter group, x (n) can be restored ).

In fact, this filter is modulated by a window function (low-pass filter) to different frequency segments to form a set of band-pass filters. The center frequency of the N channels in the filter group is the sampling value of the discrete frequency W (K) = 2PI/N * K.

The input signal passes through the filter group and corresponds to the analysis process of the phase acoustic generator. The output of each channel can be seen as a complex sine wave. This set of sine waves are superimposed or processed (such as variable speed or variable adjustment), and finally restored to the initial signal or the desired signal. It is also the basic idea of the sine model.

One of the traditional methods for Variable Speed of phase sound is based on the OLA algorithm. Considering the phase, the reconstruction of the phase is better than the OLA speech quality.

OLA variable-speed phase Acoustic Variable-speed

 

Original audio: http://pan.baidu.com/s/1i3437FJ test.wav

OLA Variable Speed 0.7: http://pan.baidu.com/s/1i3437FJ faster.wav

Phase acoustic clock speed 0.7: http://pan.baidu.com/s/1i3437FJ result.wav