pysdif3 0.5.2

Wrapper for the SDIF library for audio analysis

SDIF for Python

• Author: Eduardo Moguillansky
• Contact: [e-mail]

This is a python wrapper to IRCAM's sdif library to read and write SDIF files. It consists of a core written in Cython and some other utilities written in Python.

---

Build

• Install the SDIF library at http://sourceforge.net/projects/sdif/files/sdif/

• python3 setup.py install

• This software is released under the GPL v3 license.

---

Introduction

Sdif files are used to store time-based analysis. A Sdif file consists of time-tagged frames, each frame consisting of one or more matrices.

Read a Sdif file, read only selected matrices

    
from pysdif import *
sdif = SdifFile("path.sdif")
# get metadata
print(sdif.get_NVTs())
for frame in sdif:
    print(frame.time, frame.signature)
    for matrix in frame:
        if matrix.signature == b'1MAT':
            print(matrix.get_data())

Write a Sdif file modifying a previous one

from pysdif import *
infile = SdifFile("source.sdif")
outfile = SdifFile("out.sdif", "w").clone_definitions(infile)
for inframe in infile:
    if inframe.signature != b'1TRC':
        continue
    with outfile.new_frame(inframe.signature, inframe.time) as outframe:
        for matrix in inframe:
            # 1TRC has columns index, freq, amp, phase
            data = matrix.get_data(copy=True)
            # modify frequency
            data[:,1] *= 2
            outframe.add_matrix(matrix.signature, data)
outfile.close()

Write a SDIF file from scratch

from pysdif import *
import numpy as np

sdif = SdifFile("rbep.sdif", "w")

# Add some metadata. This is optional
sdif.add_NVT({'creator': 'pysdif3'})

# Add any matrix definitions. In this case we add only one definition
# This is a matrix named "RBEP" with 6 columns
# Each row in this matrix represents a breakpoint within a frame
# Index: partial index to which a breakpoint belongs
# Frequency: the freq. of the breakpoint
# Amplitude: the amplitude of the breakpoint
# Phase: the phase
# Bandwidth: the "noisyness" of the breakpoint
# Offset: the time offset in relation to the frame time
sdif.add_matrix_type("RBEP", "Index, Frequency, Amplitude, Phase, Bandwidth, Offset")

# After all matrix types are defined we define the frames. A frame is defined
# in terms of the matrices it accepts.
# Here we define a frame named "RBEP" which takes only matrices of type "RBEP"
sdif.add_frame_type("RBEP", ["RBEP ReassignedBandEnhancedPartials"])

# Now we need to add the data. Since there is just one matrix per frame
# in this sdif we can use the shortcut sdif.new_frame_one_matrix which 
# creates a frame and adds a matrix all at once
# The data is just fake data for the sake of an example
data = np.array([
    [1, 440, 0.1, 0, 0, 0],
    [2, 1000, 0.2, 0, 0, 0], 
], dtype=float)
sdif.new_frame_one_matrix(frame_sig="RBEP", time=0.5, matrix_sig="RBEP", data=data)

# A second frame
data = np.array([
    [1, 442, 0.1, 0, 0, 0],
    [2, 1100, 0.1, 0, 0, 0]
], dtype=float)
sdif.new_frame_one_matrix(frame_sig="RBEP", time=0.6, matrix_sig="RBEP", data=data)

sdif.close()

---

Documentation

https://pysdif3.readthedocs.io/