.wav-Classification

Classification of .wav audio file using methods for predicting interference like klt, klt_jabloun etc.

Libraries

• Pandas
• sklearn
  • Pre-processing
  • Linear Model
  • Ensemble
  • SVM
  • Metrics
  • Train Test Split

To run

Execute:

• create_dataset_from_original_file.py
• predicted_using_created_dataset.py

Improving Code

The scripts create_dataset_from_original_file.py & predicted_using_created_dataset.py are for basic understanding of Data Wrangling and Pre-Processing.

If you want higher accuracy and have knowledge of python then execute only class_predict_updated.py.

Dataset

data_svm_org_new_v2.csv contains 1793*2 data, varying between 1 - 5.

There are 8 target classes:

• babble_sn5
• car_sn5
• street_sn5
• train_sn5
• babble_sn10
• car_sn10
• street_sn10
• train_sn10.

Each class has 16 samples:

• sp01
• sp02
• sp03
• sp04
• sp06
• sp07
• sp08
• sp09
• sp11
• sp12
• sp13
• sp14
• sp16
• sp17
• sp18
• sp19 (05,10,15 is not there).

Data Wrangling

Data in column[0] is wrangled by delimiter "_".

Splitting data by "_" we create 14 columns that have methods to determine .wav file.

End product of data wrangling is to convert 1793 * 2 into 128 * 14.

Machine Learning Models

• Logistic Regression

• Random Forest Classifier

• SVC

Performance Metric

Accuracy Score