Reading MATLAB .mat Files in Python

Introduction

MATLAB is widely used for numerical computing and data analysis, often resulting in datasets saved as .mat files. If you’re working with Python but need to access these datasets, it’s essential to know how to read .mat files efficiently. This tutorial covers various methods for reading MATLAB .mat files using Python, leveraging libraries such as scipy, h5py, mat4py, and pymatreader. We’ll explore the steps needed to handle different versions of .mat files, ensuring compatibility with your datasets.

Understanding .mat Files

MATLAB files can vary in format depending on their version:

• Version 4 and 5: These are older formats that store data as MATLAB structures. They are typically easy to read using scipy.io.

• Version 7.3 (HDF5): This format uses the Hierarchical Data Format version 5 (HDF5), which requires specific libraries like h5py for access.

Reading .mat Files with SciPy

For .mat files in version 4 or 5, the scipy.io module is often sufficient:

1. Installation: Ensure you have SciPy installed:

   pip install scipy
   

2. Reading a .mat File:

   import scipy.io
   
   mat = scipy.io.loadmat('file.mat')
   print(mat)
   

3. Saving a .mat File: If you need to save data back into a .mat file, use savemat:

   scipy.io.savemat('output_file.mat', {'data': your_data})
   

4. Version Compatibility: For version 7 files, saving them as -v7 ensures compatibility.

Handling HDF5 .mat Files with h5py

For .mat files in the HDF5 format (version 7.3), use h5py:

1. Installation:

   pip install h5py
   

2. Reading a .mat File:

   import numpy as np
   import h5py
   
   with h5py.File('somefile.mat', 'r') as f:
       data = np.array(f['data/variable1'])
       print(data)
   

Using mat4py for Simple Access

mat4py offers a straightforward interface:

1. Installation:

   pip install mat4py
   

2. Loading Data:

   from mat4py import loadmat
   
   data = loadmat('datafile.mat')
   print(data)
   

3. Saving Data:

   from mat4py import savemat
   
   savemat('output_data.mat', {'key': your_data})
   

Using pymatreader for Advanced Struct Handling

pymatreader simplifies accessing structured data in MATLAB files:

1. Installation:

   pip install pymatreader pandas
   

2. Reading and Accessing Data:

   from pymatreader import read_mat
   import pandas as pd
   
   data = read_mat('matlab_struct.mat')
   keys = data.keys()
   print(keys)
   
   my_df = pd.DataFrame(data['data_opp'])
   print(my_df)
   

Conclusion

Understanding the format of your .mat file is crucial in selecting the right tool for reading it. Whether using scipy.io, h5py, mat4py, or pymatreader, each library has its strengths and can be chosen based on your specific needs, such as ease of use or handling complex structures. With these tools, integrating MATLAB datasets into Python workflows becomes seamless.