Description: 用S函数编写的量化器
量化步长为非线性
即 量化步长可调
-Prepared by S function to quantify the quantizer step size for the non-linear quantization step size that is adjustable Platform: |
Size: 5120 |
Author:天行健 |
Hits:
Description: 一种代码激励线性预测编码器，其特征在于：具有声道信息生成器、声音功率量化器、自适应码表-A code excited linear predictive coding device, characterized in that: a channel information generator, the sound power quantizer, adaptive code table Platform: |
Size: 4096 |
Author:王凤 |
Hits:
Description: 一。产生长度为500的零均值，单位方差的高斯随机变量序列，用均匀pcm的方法用16电平进行量化：1）求所得的SQNR，该序列的前5个值，相应的量化值和相应的码字。2）画出量化误差（定义为输入值和量化值之间的差），同时 画出量化值作为输入值的函数的图。3）用128量化电平数重做2）题， 比较结果。
二。产生一个长度为500，按N(O,1)分布的随机变量序列，分别用16，128量化电平数和u=255的u律非线性进行量化，画出每种情况下量化器的误差和输入-输出关系，并求SQNR.
三。长度为500的非平稳序列a由两部分组成：前20个样本是按照均值为零和方差为400的高斯随机变量产生的，其余480个样本是根据均值和方差为1的高斯随机变量产生的，对这个序列分别用均匀pcm和非均匀pcm方法进行128电平量化，试比较两种情况下所得到的SQNR。
-One. Have a length of 500 zero mean, unit variance Gaussian random variables with uniform pcm way to quantify the level with a 16: 1) Find the income SQNR, the sequence of the first 5 values, the corresponding quantization value and the corresponding codeword. 2) Draw the quantization error (defined as the input value and quantify the difference between the values), and draw quantitative values as a function of input graph. 3) redo 128 the number of quantization level 2) title, compare the results.
II. Produce a length of 500, according to N (O, 1) distributed random variables, respectively 16,128 and the number of quantization level of u u = 255 to quantify non-linear law, draw each case quantizer error and input- output relationships, and seek SQNR.
III. Length of the non-stationary series is a 500 consists of two parts: the first 20 samples in accordance with zero mean and variance of 400 generated Gaussian random variable, and the remaining 480 samples are based o Platform: |
Size: 4096 |
Author:sun |
Hits:
Description: For a file(.wave)
• Find the sampling frequency and the number of bits per sample.
• Re-quantize the samples using the following methods:
o Linear Quantization
o A-Law Companding with A=87.6
o μ-Law Companding with μ=255
• Use a mid-rise quantizer with 4, 5, 6, 7, and 8 bits per sample.
• Obtain the signal-to-quantization noise ratio (SQNR) for each of the
above 15 cases.
• It is required to plot the SQNR (dB) versus the number of bits per
sample for linear quantization, A-law companding, and μ-law
companding. All 3 plots may be on the same figure, if convenient.
-For a file(.wave)
• Find the sampling frequency and the number of bits per sample.
• Re-quantize the samples using the following methods:
o Linear Quantization
o A-Law Companding with A=87.6
o μ-Law Companding with μ=255
• Use a mid-rise quantizer with 4, 5, 6, 7, and 8 bits per sample.
• Obtain the signal-to-quantization noise ratio (SQNR) for each of the
above 15 cases.
• It is required to plot the SQNR (dB) versus the number of bits per
sample for linear quantization, A-law companding, and μ-law
companding. All 3 plots may be on the same figure, if convenient.
Platform: |
Size: 1024 |
Author:mshmsha |
Hits:
Search - linear quantizer