#!/usr/bin/env python
# coding: utf-8

# In[ ]:


import numpy, scipy, matplotlib.pyplot as plt, sklearn, librosa, mir_eval, urllib, IPython.display, stanford_mir
plt.rcParams['figure.figsize'] = (14,5)


# [← Back to Index](index.html)

# # Exercise: Instrument Classification using K-NN

# This exercise is loosely based upon "Lab 1" from previous MIR workshops ([2010](https://ccrma.stanford.edu/workshops/mir2010/Lab1_2010.pdf)).

# For more on K-NN, see the [notebook on K-NN](knn.ipynb).

# For help from a similar exercise, [follow the steps in the feature sonification exercise](feature_sonification.ipynb#Step-1:-Retrieve-Audio) first.

# ## Goals

# 1. Extract spectral features from an audio signal.
# 2. Train a K-Nearest Neighbor classifier.
# 3. Use the classifier to classify beats in a drum loop.

# ## Step 1: Retrieve Audio, Detect Onsets, and Segment

# Download the file `simple_loop.wav` onto your local machine.

# In[ ]:


filename = 'simple_loop.wav'
get_ipython().run_line_magic('pinfo', 'urllib.urlretrieve')


# Load the audio file:

# In[ ]:


get_ipython().run_line_magic('pinfo', 'librosa.load')


# Play the audio file:

# In[ ]:


get_ipython().run_line_magic('pinfo', 'IPython.display.Audio')


# Detect onsets:

# In[ ]:


get_ipython().run_line_magic('pinfo', 'librosa.onset.onset_detect')


# Convert onsets from frames to seconds (and samples):

# In[ ]:


get_ipython().run_line_magic('pinfo', 'librosa.frames_to_time')


# In[ ]:


get_ipython().run_line_magic('pinfo', 'librosa.frames_to_samples')


# Listen to a click track, with clicks located at each onset, plus the original audio:

# In[ ]:


get_ipython().run_line_magic('pinfo', 'mir_eval.sonify.clicks')


# In[ ]:


get_ipython().run_line_magic('pinfo', 'IPython.display.Audio')


# ## Step 2: Extract Features

# For each segment, compute the zero crossing rate and spectral centroid.

# In[ ]:


get_ipython().run_line_magic('pinfo', 'librosa.zero_crossings')


# In[ ]:


get_ipython().run_line_magic('pinfo', 'librosa.feature.spectral_centroid')


# Scale the features to be in the range [-1, 1]:

# In[ ]:


get_ipython().run_line_magic('pinfo', 'sklearn.preprocessing.MinMaxScaler')


# In[ ]:


get_ipython().run_line_magic('pinfo', 'sklearn.preprocessing.MinMaxScaler.fit_transform')


# ## Step 3: Train K-NN Classifier

# Use `stanford_mir.download_drum_samples` to download ten kick drum samples and ten snare drum samples. Each audio file contains a single drum hit at the beginning of the file.

# In[ ]:


get_ipython().run_line_magic('pinfo', 'stanford_mir.download_drum_samples')


# For each audio file, extract one feature vector. Concatenate all of these feature vectors into one feature table.

# In[ ]:


get_ipython().run_line_magic('pinfo', 'numpy.concatenate')


# ## Step 4: Run the Classifier

# Create a K-NN classifer model object:

# In[ ]:


get_ipython().run_line_magic('pinfo', 'sklearn.neighbors.KNeighborsClassifier')


# Train the classifier:

# In[ ]:


get_ipython().run_line_magic('pinfo', 'sklearn.neighbors.KNeighborsClassifier.fit')


# Finally, run the classifier on the test input audio file:

# In[ ]:


get_ipython().run_line_magic('pinfo', 'sklearn.neighbors.KNeighborsClassifier.predict')


# ## Step 5: Sonify the Classifier Output

# Play a "beep" for each detected kick drum. Repeat for the snare drum.

# In[ ]:


get_ipython().run_line_magic('pinfo', 'mir_eval.sonify.clicks')


# ## For Further Exploration

# In addition to the features used above, extract the following features:
# 
# - spectral centroid
# - spectral spread
# - spectral skewness
# - spectral kurtosis
# - spectral rolloff
# - MFCCs

# Re-train the classifier, and re-run the classifier over the test audio signal. Do the results change?

# Repeat the steps above for more audio files.

# [← Back to Index](index.html)