citwild / laugh-finder Goto Github PK

View Code? Open in Web Editor NEW

23.0 4.0 3.0 56 KB

A machine learning program for human laughter detection.

Java 48.66% Python 51.34%

laugh-finder's Introduction

Laugh Finder

A machine learning program for human laughter detection.

Setup

Install Python 3.4.3 or later
Install Python's SciPy package and related NumPy package

If using Windows, access NumPy and SciPy wheels here
- Note your version of Python and architecture (i.e., 32-bit v. 64-bit)

Install WEKA

Using the Programs

Training

Generate .arff File for WEKA

Run main.py under the python-training folder
In the GUI, click "Browse" next to "File with laughter audio file names"
Select the file containing a series of paths to laughter training files
Click "Browse" next to "File with non-laughter audio file names"
Select the file containing a series of paths to non-laughter training files
Click "Prepare" to generate a wekaFile.arff file within the same directory as main.py

Using WEKA

Run WEKA using the following command:

java -Xmx512m -classpath /location/of/weka/jar/weka.jar weka.gui.GUIChooser

In the GUI, click the "Explorer" button
Under the "Preprocess" tab, click the "Open file..." button
Select the wekaFile.arff file generated earlier
Under the "Classify" tab, select "Choose" under "Classifier"
To use K-nearest neigbors algorithm: under "lazy", select "IBk"
Press "Start"
After the classification is finished, right-click the output in "Result List" and click "Save Model"

Testing

Run main.py under the python-testing folder

Include the following params:
- --phase 0; for testing algorithm
- --audio /path/to/audio.wav; the audio file you want to examine
  - .wav files only, not 24 bit depth (limitation of SciPy)
- --arff /path/to/arff/result.arff; the result .arff file from training

Re-Training

Change file paths in Constants.java accordingly, relative to the values of your machine
Compile and run the WavReader main class (include the weka.jar to your classpath)
Run the main Java file
In the GUI, select the audio file containig the laughter segments to be identified
Click the "Get Laughter Segments" button
You will be presented with a series of snippets identified as laughter, select each one that is correctly identified
Click the retrain model process
The re-trained model will be placed in the same directory as the old model (see Constants.java)

Useful Commands

ffmpeg

If you are using a Macintosh, you can install ffmpeg using Homebrew

brew install ffmpeg

Using that application, you can do the following:

Split a wav file into segments
- ffmpeg -i audio.wav -f segment -segment_time 0.5 -c copy parts/out%03d.wav
Find the exact duration of a wav file
- ffmpeg -i audio.wav 2>&1 | grep Duration | awk '{print $2}' | tr -d ,
Cut for a given duration of a wav file
- ffmpeg -i audio.wav -ss 00:00:07.100 -t 00:00:01.1 audio.wav
Convert .mp4 to .wav
- ffmpeg -i sourcefile.mp4 destination.wav

Reading

laugh-finder's People

Contributors

Stargazers

Watchers

Forkers

ohkhan zhanghonglishanzai fthornton67

laugh-finder's Issues

Provide automated ripping audio from .mp4 video files

Symbol not found: ___addtf3

Hi Guys,

Anyone seen this error ? How did u fix it?

ImportError: dlopen(/Users/rock/laugh/laugh-finder/python-training/laughenv/lib/python3.6/site-packages/scipy/linalg/_fblas.cpython-36m-darwin.so, 2): Symbol not found: ___addtf3
Referenced from: /Users/rock/laugh/laugh-finder/python-training/laughenv/lib/python3.6/site-packages/scipy/linalg/../.dylibs/libquadmath.0.dylib
Expected in: /usr/local/lib/libgcc_s.1.dylib
in /Users/rock/laugh/laugh-finder/python-training/laughenv/lib/python3.6/site-packages/scipy/linalg/../.dylibs/libquadmath.0.dylib
(laughenv) rockmac:python-training rock$ python3 main.py
Traceback (most recent call last):
File "main.py", line 6, in
from file_feature_extraction import file_feature_extraction
File "/Users/rock/laugh/laugh-finder/python-training/file_feature_extraction.py", line 1, in
import scipy.io.wavfile as wavfile
File "/Users/rock/laugh/laugh-finder/python-training/laughenv/lib/python3.6/site-packages/scipy/io/init.py", line 97, in
from .matlab import loadmat, savemat, whosmat, byteordercodes
File "/Users/rock/laugh/laugh-finder/python-training/laughenv/lib/python3.6/site-packages/scipy/io/matlab/init.py", line 13, in
from .mio import loadmat, savemat, whosmat
File "/Users/rock/laugh/laugh-finder/python-training/laughenv/lib/python3.6/site-packages/scipy/io/matlab/mio.py", line 12, in
from .miobase import get_matfile_version, docfiller
File "/Users/rock/laugh/laugh-finder/python-training/laughenv/lib/python3.6/site-packages/scipy/io/matlab/miobase.py", line 22, in
from scipy.misc import doccer
File "/Users/rock/laugh/laugh-finder/python-training/laughenv/lib/python3.6/site-packages/scipy/misc/init.py", line 53, in
from scipy.interpolate._pade import pade
File "/Users/rock/laugh/laugh-finder/python-training/laughenv/lib/python3.6/site-packages/scipy/interpolate/init.py", line 176, in
from .interpolate import *
File "/Users/rock/laugh/laugh-finder/python-training/laughenv/lib/python3.6/site-packages/scipy/interpolate/interpolate.py", line 20, in
import scipy.linalg
File "/Users/rock/laugh/laugh-finder/python-training/laughenv/lib/python3.6/site-packages/scipy/linalg/init.py", line 175, in
from .misc import *
File "/Users/rock/laugh/laugh-finder/python-training/laughenv/lib/python3.6/site-packages/scipy/linalg/misc.py", line 5, in
from .blas import get_blas_funcs
File "/Users/rock/laugh/laugh-finder/python-training/laughenv/lib/python3.6/site-packages/scipy/linalg/blas.py", line 155, in
from scipy.linalg import _fblas
ImportError: dlopen(/Users/rock/laugh/laugh-finder/python-training/laughenv/lib/python3.6/site-packages/scipy/linalg/_fblas.cpython-36m-darwin.so, 2): Symbol not found: ___addtf3
Referenced from: /Users/rock/laugh/laugh-finder/python-training/laughenv/lib/python3.6/site-packages/scipy/linalg/../.dylibs/libquadmath.0.dylib
Expected in: /usr/local/lib/libgcc_s.1.dylib
in /Users/rock/laugh/laugh-finder/python-training/laughenv/lib/python3.6/site-packages/scipy/linalg/../.dylibs/libquadmath.0.dylib

Thanks
Rock

Playback of found laughter segments don't work when player is stopped

Not a big issue, since the UI is expected to be replaced with a web interface (ideally); just leaving as a note for others if they choose to use it..

ValueError: embedded null character

Try to prepare files with README:

Run main.py under the python-training folder

In the GUI, click "Browse" next to "File with laughter audio file names"

Select the file containing a series of paths to laughter training files

Click "Browse" next to "File with non-laughter audio file names"

Select the file containing a series of paths to non-laughter training files

Click "Prepare" to generate a wekaFile.arff file within the same directory as main.py.

But I have an error:

File "C:\Users\...\Python\Python37\lib\site-packages\scipy\io\wavfile.py", line 544, in read
    fid = open(filename, 'rb')
ValueError: embedded null character.

How can I solve this problem?