Spoken Language Toolkit
This library provides simple access to many common Praat speach feature information, and a wrapper around
the python_speech_features library from https://github.com/jameslyons/python_speech_features.
Installation

This project is on pypi

To install from pypi:

pip install sltk

From this repository:

git clone https://github.com/nickdepinet/sltk
python setup.py develop

Usage

Supported features:

• Pitch
• Intensity
• Formants
• Harmonicity

Features from python_speech_features

• Mel Frequency Cepstral Coefficients
• MFCC deltas
• Filterbank Energies
• Log Filterbank Energies

Initialization

You need to initialize the SLTK object with the path of your Praat binary:

python
from sltk import SLTK
sltk = SLTK("/path/to/praat")

If Praat is already in your path, you can leave this blank:

python
sltk = SLTK()

For these functions, I've provided the default parameters as Praat provides them, so you can usually just run with the bare minimum required arguments and get what you want

Pitch Features

You can easily get the pitch values over an interval of time. The default start and end give the values for the whole file:

python
min, mean, max, stdev = sltk.pitch_interval(wavfile, time_step=0, pitch_floor=75, pitch_ceil=600, start_time=0, end_time=0)

Parameter	Description
wavfile	the wavfile to read from
time_step	the window size, default is a multiple of the pitch floor
pitch_floor	the lowest pitch to consider
pitch_ceil	the highest pitch to consider
start_time	the start of the analysis time (default is the whole file)
start_time	the end of the analysis time (default is the whole file)
returns	the minimum, mean, maximum, and standard deviation for pitch during the interval

If you want a pitch value at a specific time, that is easy to do as well. This will return -1 if the pitch at the specified time is undefined:

python
pitch_val = sltk.pitch(wavfile, pitch_time, time_step=0, pitch_floor=75, pitch_ceil=600)

Parameter	Description
wavfile	the wavfile to read from
pitch_time	the time to get the pitch at
time_step	the window size, default is a multiple of the pitch floor
pitch_floor	the lowest pitch to consider
pitch_ceil	the highest pitch to consider
returns	the pitch value at the specified time, or -1 if the pitch is undefined (the frame is unvoiced)

Formant Features

Most of the time, you just want a formant or formants at a specific time in the file:

python
f1 = sltk.f1(wavfile, formant_time)
f2 = sltk.f2(wavfile, formant_time)
f3 = sltk.f3(wavfile, formant_time)
f4 = sltk.f4(wavfile, formant_time)
f1, f2, f3, f4 = sltk.all_formants(wavfile, formant_time)

Parameter	Description
wavfile	the wavfile to read from
formant_time	the time to get the formant at
returns	the formant at the specified time

In some cases, you want more detailed information than this. In these cases, you can use the function that all of those wrap:

python
formant_val = sltk.formant(wavfile, formant, formant_time, time_step=0, max_num_formants=5,
                        max_formant=5500, window_length=0.025, pre_emphasis=50))

Parameter	Description
wavfile	the wavfile to read from
formant	the formant number to get
formant_time	the time to get the formant at
time_step	used by Praat to calculate the window size
max_num_formants	the highest formant to find
max_formant	the highest Frequency to considered a formant at
window_length	the length of the windows for calculation
pre_emphasis	the pre_emphasis strength to apply before calculation
returns	the formant at the specified time

MFCC Features

The default parameters should work fairly well for most cases, if you want to change the MFCC parameters, the following parameters are supported:

python
mfcc_arr = sltk.mfcc(wavfile,winlen=0.025,winstep=0.01,numcep=13,
        nfilt=26,nfft=512,lowfreq=0,highfreq=None,preemph=0.97,
        ceplifter=22,appendEnergy=True)

Parameter	Description
wavfile	the wavfile to read from
winlen	the length of the analysis window in seconds. Default is 0.025s (25 milliseconds)
winstep	the step between successive windows in seconds. Default is 0.01s (10 milliseconds)
numcep	the number of cepstrum to return, default 13
nfilt	the number of filters in the filterbank, default 26.
nfft	the FFT size. Default is 512
lowfreq	lowest band edge of mel filters. In Hz, default is 0
highfreq	highest band edge of mel filters. In Hz, default is samplerate/2
preemph	apply preemphasis filter with preemph as coefficient. 0 is no filter. Default is 0.97
ceplifter	apply a lifter to final cepstral coefficients. 0 is no lifter. Default is 22
appendEnergy	if this is true, the zeroth cepstral coefficient is replaced with the log of the total frame energy.
returns	A numpy array of size (NUMFRAMES by numcep) containing features. Each row holds 1 feature vector.

MFCC Deltas

Using the mfcc array returned by the above function, you can pretty easily get the deltas:

python
deltas = sltk.deltas(mfcc_arr, distance)

Parameter	Description
mfcc_arr	the mfcc numpy array
distance	the distance in either direction to compare
returns	A numpy array of the mfcc deltas

Filterbank Features

These filters are raw filterbank energies. For most applications you will want the logarithm of these features. The default parameters should work fairly well for most cases. If you want to change the fbank parameters, the following parameters are supported:

python
fbank_arr = sltk.fbank(wavfile,winlen=0.025,winstep=0.01,
        nfilt=26,nfft=512,lowfreq=0,highfreq=None,preemph=0.97)

Parameter	Description
wavfile	the wavfile to read from
winlen	the length of the analysis window in seconds. Default is 0.025s (25 milliseconds)
winstep	the step between seccessive windows in seconds. Default is 0.01s (10 milliseconds)
nfilt	the number of filters in the filterbank, default 26.
nfft	the FFT size. Default is 512.
lowfreq	lowest band edge of mel filters. In Hz, default is 0
highfreq	highest band edge of mel filters. In Hz, default is samplerate/2
preemph	apply preemphasis filter with preemph as coefficient. 0 is no filter. Default is 0.97
returns	A numpy array of size (NUMFRAMES by nfilt) containing features. Each row holds 1 feature vector. The second return value is the energy in each frame (total energy, unwindowed)