WEEK 2 — MUSIC GENRE CLASSIFICATION
Hello, this is our second blog post for our Fundamentals of Machine Learning course project. This week, we have decided on our data set and format for music genre classification. So let’s begin!
This week, for our aim to make a study that predicts the type and emotion of the music, we decided on our dataset. This dataset is a collection of 10 genres with 100 audio files each, all having a length of 30 seconds. We will use this dataset for classifying the genres of songs. Also, we want to predict song emotions. For this purpose, we decided to use song texts. Generally, we will train our models with two datasets, and as a result, we will predict the genre and emotion of one given song with high accuracy.
We will use wav (Waveform Audio File) format sound files. These files represented in the form of audio signals and these signals have different kinds of features such as frequency, bandwidth, decibel, etc. We can represent the audio file with Librosa library and it looks like
In our project, we will extract the characteristics that are relevant to our classification problem. The process of extracting features to use them for analysis is called feature extraction. After extracting, we will manipulate our extracted data to make it more suitable for our problem. In this way, our prediction will be more desirable. After that, we will try to model a classifier to classify songs into different genres.
DATASET
NEXT STEP
Genre Classification is one of the branches of Music Information Retrieval and is a very popular branch. In the next post, we will try to give more detailed explanations about algorithms that will be used in our project.
