mcrblg-header-image

search

RNN vs CNN in Speech Recognition

Published on 2023-10-06 in Speech Recognition

RNN or CNN, that’s the question. so what should you use?

Let the battle begin

So speech recognition is a very broad task. People use speech recognition to do speech-to-text on videos, a pre-recorded data which you can go back and forth between past and future and optimize the output a couple of times. On the other hand voice control is also a speech recognition task, but you need to do all this speech processing in real time. And within a low latency time manner.

And now comes the big question. Which technologies should you use RNN or CNN? in this post, We’re going to talk about that

so generally, speech waveform data by using a CMVN or MFCC, can be converted to 2D image data and then, from that point is basically an image that you can show to people and people can learn how each word will look like. So, it is basically detecting where exactly the word is happening. And it’s very similar to an object detection task. very similar, but not quite the same. And why is that? So, a lot of times we also have trouble detecting words but we are using the language grammar in the background of our head to predict what exactly the next word would look like. And we’re using that and combining that with the waveform data and then we detect the right word so if you say, a very strange word to people, they will have trouble getting the correct text out of it. But if you teach them a couple of times, and they know that when these words pop up, they will have much less trouble detecting them. So the machine learning community uses the same approach.

In modern speech recognition engines ML engineers first use CNN to capture the features, or at least detect how likely the word is, and then run an RNN in the background as a language model to improve the result. So in the case of voice control, you don’t care about the language model, because there is no language model. You can say whatever word you want, or at least we give you that freedom, and then you just need the text out of the word that you just said.

So in that case there is no use for RNN as there is no language model. And a 1D convolutional network is enough. So it is the same as localization and object detection in classic machine learning. So if you use the same technique as YOLO to move around the convolution layer, around the waveform, and just detect the maximum confidence score on a window, then you can find the exact word happening at that time. The problem is as the number of words increases and increases this technique will become more and more challenging. So you need to develop more mature techniques. And that’s exactly why we introduced HMM. The best technique is to use a hidden Markov model To detect which word is spoken in a certain way and then slide that word over the signal and find out if it’s actually that signal or not. And by using that we can do that alignment. We can do better force alignment, use that data, and also feed it to Hmm, To increase the accuracy and finally, we create this awesome engine with a great amount of accuracy that no one has seen ever before.

So wait for it and Sleep on it.


Windows Accurate Timing

Published on 2023-09-13 in Software, Windows

There’s a lot of controversial discussion over the internet about achieving accurate sleep function on Windows platforms. The problem is most of them are very old and with the introduction of multicore processors many of the older functions break down but this is not the case in 2023.
Nowadays you can easily call the cross-platform C++ 11 chrono function and with the following source code, I could achieve one millisecond accuracy which is more than enough for my application.

#include <chrono>
#include <thread>
std::this_thread::sleep_for(std::chrono::microseconds(500));

Before I used the QThread::msleep function which had an accuracy of about 5ms to 15ms which was a lot more than what I imagined even when I used the QThread::usleep function.
There is an issue that is if you call the sleep function on a thread the OS scheduling system will put your application to sleep and it may take a while till the scheduler picks up your application again. To prevent this issue you need to specifically tell OS to treat your application differently than others and C++ 11 introduces chrono which uses QueryPerformanceCounter in the background Windows API to make sure Windows scheduler will pick your application up at the right time
You can go ahead and directly call the Windows API function but nowadays C++ 11 is nicely integrated into a lot of environments and it’s also a cross-platform solution so lucky you, you don’t need to get your hands dirty anymore.


  1. YouTube – Test and Set Synchronization Primitive
  2. RandomASCii: Windows Timer
  3. Microsoft – Windows Performance Analyzer
  4. YouTube – CppCon 2017: Fedor Pikus “C++ atomics, from basic to advanced.”

Windows Last Restart Reason

Published on 2023-09-11 in Windows

The Windows 10 may restart unexpectedly. In order to find out why you can execute the following command:

wevtutil qe System /q:"*[System[(EventID=41) or (EventID=1074) or (EventID=6008)]]" /c:1 /f:text /rd:true

If the case is Operating System: Service pack (Planned), set the following registry key to disable automatic Windows update restart:

HKLM\SOFTWARE\Policies\Microsoft\Windows\WindowsUpdate\AU\NoAutoRebootWithLoggedOnUsers

SetWindowsHookEx Callback Never Called

Published on 2023-08-05 in Software, Windows

Previously, we learned we needed to put our callback function in a separate DLL and then load that DLL in another program. So you did that you create a separate dll, you put the callback there you call the SetWindowsHookEx function to set up the callback, But the callback doesn’t get called at all. So what’s wrong?

In order to figure this out, you need to notice that the SetWindowsHookEx function is actually a DLL injector. What does that mean? It means it first loads or injects your DLL inside the target application, And then it will hook your callback to the Windows messages queue or any other callback that you specified.

So for this to work, there are several conditions that need to be met otherwise several things can go wrong. And in my case, all of them went wrong. I’m going to list them here. So maybe it will help you. By any chance if you bump into any other issue and you finally got to solve it. Then please mention that in the comments, so other people and I can learn from your mistakes as well. Alright, let’s get to the problems.

1. Thread ID

The first problem that I had was the thread ID. So there are several ways to get the thread ID. First, don’t use the GetWindowThreadProcessId function

the best and most legitimate way to get the thread ID is by using the CreateToolhelp32Snapshot function which goes through all threads and checks if they are the main thread for an application. Actually, I found that the CreateToolhelp32Snapshot function is returning a different value than the GetWindowThreadProcessId, and it was actually the reason why my DLL doesn’t get floats. So that’s first.

2. DLL Dependency

Second, you have to be aware that your callback is going to inject your DLL into another program. So you have to make sure that the DLL that you wrote is not dependent on any other libraries other than the one that is already installed in the “System Path”.

so in my case, I was using the Qt library and the SetWindowsHookEx function was installing the hook successfully, But in reality, the dll never gets injected into the target program.

3. Printf

The last dilemma was, don’t use the printf function without taking extra care. So in my case, I was using this

To know why you got to understand that when you’re calling this function from your application and then you call AllocConsole function, it will open up a console that is hooked up to your application not to the DLL that is injected into the program so if you printf something in your callback, you won’t see anything because it’s running on a thread on your main application. If you want to see that you need to call the AllocConsole function inside your DLL main function.

Otherwise when you use the printf function that will write to the output of the console of the target processor instead of the console that you had opened


What do the letters W and L stand for in WPARAM and LPARAM?


SetWindowsHookEx without DLL?

Published on 2023-07-27 in Software, Windows

Previously we discussed why it isn’t required to inject the DLL containing the callback to every application you need to set a hook on. In this post, we are going to see if it’s possible to create a hook or a global Hook with SetWindowsHookEx function without using any sort of DLL.
There is this cool trick in which you can use some known Dll to grab their handle and use that to legitimately install the hook with this function. Here is the code for doing that

HMODULE dll_handle = GetModuleHandle(L"user32.dll");
windows_exhook = SetWindowsHookExA(WH_GETMESSAGE, CallWndProcHook,
							dll_handle, tid); //0: System wide event
DWORD error = GetLastError();
if( windows_exhook==NULL )
{
	qDebug() << "Failed to set the hook. Error code: "
			 << error;
}
qDebug() << "hook was successfully installed="
         << windows_exhook << dll_handle
         << error << tid;

Although this code runs successfully and installs the hook if you install this globally by setting the thread id to 0, it will Instantly crash your whole explorer processor.
If you try that on a single application it also crashes the application.

I tried and spend some time to figure this out and it turns out that the callback will never get executed. I tried to simplify the callback do some sanity tests and it looks like the hook is installed without any problems and sometimes the callback called and most of the time it won’t but at the end in all test cases the targeted application were always crashes. Here is the simplified callback function that I used:

LRESULT CALLBACK CallWndProcHook(int nCode, WPARAM wParam,
                                 LPARAM lParam)
{
    // Call the next hook in the chain
    // The first argument is ignored so any value will work
    return CallNextHookEx(windows_exhook, nCode,
                          wParam, lParam);
}

I searched a bit about this and there were two comments around this issue

  1. The first was that this is because you are skipping the phase which is putting your callback inside a Dll [link]
  2. The second one is it’s because of some sort of aggressive security application that detects that this is not a legitimate behavior so it will instantly kill the effected affected application in order to prevent any data hijacking [link]

Conclusion

Although at the beginning it seems that you can get away using SetWindowsHookEx function without a Dll currently at least up to what I know there is no possible way to use SetWindowsHookEx function without having the callback inside another Dll


SetWindowsHookEx with WH_GETMESSAGE

Published on 2023-07-05 in Windows

There was this special application that I wanted to intercept all Windows messages for a specific remote processor. For that, I came to SetWindowsHookEx function but in order to install this hook on an external application you cannot do that simply by calling this function inside your code because it will fail with the following error

ERROR_HOOK_NEEDS_HMOD
1428 (0x594)
Cannot set nonlocal hook without a module handle.

In order to fix that at first I thought I need to create a DLL which contains the hook function and then inject that into every application that I need to intercept its messages. What I later found out was that this is not required
To shed some light in there there are two subjects around this, first, there is this DLL injection topic which you can use this function to inject some DLL into other programs and the second one is my application which intercepts events from other applications. For the first one you need to inject the dll which contains the codes you want to inject into the application you want, obviously.
But for the second application that I mentioned you just need to write your codes inside some dll and then just load that codes inside your application and because you wrote down the SetWindowsHookEx function and the hook callback inside the DLL main function your code will successfully execute and actually you don’t need to inject it into any other application so it is pretty straightforward.
Because at first, it seems to be so complicated I decided to write this down so I wouldn’t forget it later.
That it, enjoy and if it was helpful please send some feedback.

P.S.: SetWinEventHook Is a completely different Hook. What exactly it does is it reports the events regarding the window manager. To be more specific it reports events related to window creation and destruction, window resizing and losing, or getting Focus around different windows. But it would never give any response regarding the window internal messages.


Mark Jansen: WindowsHookEx GitHub Project

Stmxcsr: Reinventing the wheel, DLL Injection via SetWindowsHookExA


Batch Size vs Mini Batch Size

Published on 2023-06-03 in Speech Recognition

slavv – 37 reasons why your neural network-is-not-working

Intel – TBB


Golden SSH Commands

Published on 2023-05-19 in Linux
  1. Are you tired of entering your password for sshing into a remote server?
  2. Are you tired of verifying host fingerprint changes every time?

Use this golden ssh command and save your valuable time

ssh -o "StrictHostKeyChecking no" -i ~/Documents/identify root@ip
echo "y\n" | HOSTNAME=`hostname` ssh-keygen -t rsa -C "$HOSTNAME" -f "$HOME/.ssh/id_rsa" -P ""
# copy key to remote
ssh-copy-id userid@hostname
sshpass -p pass ssh root@ip

CronTab @Reboot

Published on 2023-05-19 in Linux

You can use the @reboot keyword in crontab to start a shell script at system startup but here is why this isn’t a very good solution to do that.

The problem is that if you don’t shut down the system cleanly on the next startup this message will pop up and cron will simply skip over running your command.

"Skipping @reboot jobs -- not system startup"

The solution is easy, just use a systemd service.

/etc/systemd/system/service_name.service
------------------------------------------
[Unit]
Description=some description
After=network.target
StartLimitIntervalSec=0

[Service]
Type=simple
User=root
ExecStart=/home/user/script.sh

[Install]
WantedBy=multi-user.target

 

UnixDaemon – How Does Cron Reboot Work


Useful Firefox/Windows Hacks

Published on 2023-02-24 in Windows

Disable Firefox Update Prompt

run in cmd

reg ADD HKLM\SOFTWARE\Policies\Mozilla\Firefox /v DisableAppUpdate /t REG_DWORD /d 1 /f

or run ff_update_dis.reg

Disable Windows Keys

reg ADD HKLM\SOFTWARE\Microsoft\Windows\CurrentVersion\Policies\Explorer /v NoWinKeys /t REG_DWORD /d 1 /f

or run win_key_dis.reg

Disable Shutdown Prompt Ending Application

reg ADD HKCU\Control Panel\Desktop /v AutoEndTasks /t REG_SZ /s "1" /f

or run win_end_task.reg

Disable Share Screen Prompt/Overlay

privacy.webrtc.hideGlobalIndicator

media.navigator.permission.disabled

Auto YouTube Play

media.block-autoplay-until-in-foreground

AddOn Restriction URL

extensions.webextensions.restrictedDomains -> set to null

privacy.resistFingerprinting.block_mozAddonManager -> true


  1. Mozilla – Firefox Marionete

‹ previous posts next posts ›
close
menu