Half plotting problem of spectrogram

Question

0 个投票

I am producing spectrograms of 10 hour audio in chunks of 4 seconds, by using the "for" loop and storing it in a folder. But many of the spectrograms generated and saved are incomplete. Why it is so? and how I can stop this to happen

This problem is not for all spectrograms, only for 30% of the total.

2 个评论
显示无隐藏无

Chunru 2022-10-27

Show your program.

Sania Gul 2022-10-27

在 MATLAB Online 中打开

Here is my code :)

clear all;
clc;
close all;
Fs=16000;
fs=Fs;
audio=[];
opening=0;
count =1;
% Load the source file
while 1
        
samples = [1+opening,count*Fs*4.12];% 16 time frames corresponds to 0.287 seconds interval
if samples(2)>540938091% total sample in audio train file
    break
end
% if samples(2)>33152737% total sample in audio test file
%     break
% end
audio=audioread('D:\Fazeel_Data\16KAll\TrainNoisy.wav',samples);
%Destination folder for storing images
Dest = 'D:\Fazeel_Data\Parula\TrainNoisyeePicsDup\';
% figure(1);
% spectrogram(audio,hann(512),256,512,Fs,'yaxis');
[Q,F,T] = spectrogram(audio,hann(512),256,512,Fs,'yaxis');
% Taking the log of Z-Axis (Brightness)
Q=Q(1:256,:);
F=F(1:256);
figure(2)
D=surf(T,F,log(abs(Q).^2));
colormap parula
shading interp
view([0 90])
axis tight
set(gca,'xtick',[])
set(gca,'ytick',[])
% print('AB','-djpeg','-r50');
% I=imread('AB.jpg');
% [a b c]=size(I)
export_fig AA.jpg  -native -c[31 52 46 72]% top right bottom left
% resizing the image
I=imread('AA.jpg');
outputImage = imresize(I, [256,256]);
% imshow(outputImage);
export_fig AA.jpg  -native -c[31 86 58 85]% top right bottom left
% imshow(outputImage);
% Copying the figure from current directory to Destination folder
% psource = 'D:\Fazeel_Data';
% pattern = 'AA.jpg';
% sourceFile = fullfile(psource, pattern);
[o p q]=size(outputImage)
imagefilename = sprintf('TrainNoisyee_%05d.jpg',count);
   
    destFile   = fullfile(Dest, imagefilename);  % Or another name?
%     copyfile(sourceFile, destFile);
    imwrite(outputImage, destFile);
    opening=samples(2);
    count=count+1;
%     pause(1);
end

end

请先登录，再进行评论。

请先登录，再回答此问题。

Follow Question

Answer 1

VBBV 2022-11-7

编辑：VBBV 2022-11-7

在 MATLAB Online 中打开

0 个投票

imwrite(outputImage, destFile,'Quality',"lossless");

It appears that some part of the image data is being lost during write operation. Try with name value argument specifying quality as lossless

6 个评论
显示 4更早的评论隐藏 4更早的评论

VBBV 2022-11-8

编辑：VBBV 2022-11-8

在 MATLAB Online 中打开

I = imread('peppers.png');

imshow(I)

[r c] = size(I) % get the size of image

r = 384

c = 1536

if r == c

outputImage = imresize(I,[r c]); % this could be reason, change using scale value

elseif r < c

outputImage = imresize(I,[r+abs(c-r) c]); ;% this could be reason, change using scale value

elseif r > c

outputImage = imresize(I,[r c+abs(c-r)]);

end

outputImage = imresize(I,[120 c]); % see the difference with fixed row and col values

imshow(outputImage)

% imwrite(outputImage, destFile,"Quality",100); % switch back to normal mode

Note the difference when you use fixed numbers for rows and cols for all image sizes, Try with the higher or scaled values for rows and cols inside the imresize function and switch back to normal mode. May be some images have high resolution with large or different sizes,

Sania Gul 2022-11-9

Tnk u VBBV. The photo is compressed, however complete in your case. However, it is half drawn or incomplete in my case. I have to be restricted in defining the image size as my spectrograms are given as input to pix2pix network. Secondly I have generated 8000 spectrograms by the same set of instructions and only random 30% suffer from this problem. This may seems to be some issue in memory allocation by the processor as C drive is almost full. But anyways thanks for your time and efforts. I regard them high :-)

请先登录，再进行评论。

Half plotting problem of spectrogram

2 个评论
显示无隐藏无

采纳的回答

6 个评论
显示 4更早的评论隐藏 4更早的评论

更多回答（0 个）

类别

标签

Community Treasure Hunt

Half plotting problem of spectrogram

2 个评论 显示 无 隐藏 无

采纳的回答

6 个评论 显示 4更早的评论 隐藏 4更早的评论

更多回答（0 个）

类别

标签

另请参阅

Community Treasure Hunt

2 个评论
显示无隐藏无

6 个评论
显示 4更早的评论隐藏 4更早的评论