还是一样，先贴程序

%sampling and quantisation

clc;

clear;

%sampling

A = 1; % sinusoid's amplitude

f = 2;% frequency Hz

phase = 0*pi;% phase Rad

f_sample = 1000;%sample frequecy

t_start = 0;

t_stop = 1;

t = t_start:1/f_sample:t_stop;

y = A*sin(2*pi*f*t+phase); % y includes all sampled value

figure(1);

plot(t,y);

xlabel('t');

grid on;

%quantisation

n = 2; % n bits represent the number of 2^n combinations

M = 2^n;% quantization level

delta_v = 2*A/M; % quantization interval

m_i = zeros(1,M);% mi = a + i*delta_v, quantization_endpoint_value

for i = 1:M

m_i(i) = -A + i*delta_v;

end

m_i_ahead = [-A m_i(1:end-1)]; % mi-1

% quantization_value qi = (mi+mi-1)/2

q_i = (m_i + m_i_ahead)/2;

% find the sampled valued which is the nearst to quantization value

y_hat = repmat(y,M,1);

q_i_hat = repmat(q_i.',1,size(y_hat,2));

nearst_value = abs(y_hat - q_i_hat);

[min_value min_location] = min(nearst_value,[],1);

min_location_hat = min_location-1;%because the value of min_location is from 1 to 8, change

% it from 0 to 7, in order to turn it to 000,001,010...

y_digital = dec2bin(min_location_hat,n);

% plot quantization_value

hold on;

plot(t,q_i_hat(min_location),'r');

legend('sampled value','quantized value')

程序的第一部分我在之前发的一个帖子里说明过，就是一个时间上离散但是数值上是连续的信号(抽样过的信号)，那么接下来我们要做的就是给信号量化成数字信号。具体的理论书上都有，不在这细谈，可以书上查找查找，我这里仿真的是最简单的均与量化，至于非均匀量化，看懂原理，花费些时间，都可以编出来。

量化电平数 M = 2^n;

均匀量化间隔 delta_v = (b-a)/M;

量化区间终点 mi = a+i*delta_v ;

量化区间中点 qi = (mi+mi_1)/2, i = 1,2,3...M,

5.jpg (42.35 KB, 下载次数: 11)

2013-12-4 22:17 上传

6.jpg (42.07 KB, 下载次数: 10)

2013-12-4 22:17 上传

7.jpg (41.87 KB, 下载次数: 11)

2013-12-4 22:17 上传

有兴趣的可以下下来试试，修改n的值就可以修改量化电平数，量化电平数越少，量化误差越大。