# ------------------------------------------------------------------------ #
# Copyright 2022 SPTK Working Group #
# #
# Licensed under the Apache License, Version 2.0 (the "License"); #
# you may not use this file except in compliance with the License. #
# You may obtain a copy of the License at #
# #
# http://www.apache.org/licenses/LICENSE-2.0 #
# #
# Unless required by applicable law or agreed to in writing, software #
# distributed under the License is distributed on an "AS IS" BASIS, #
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. #
# See the License for the specific language governing permissions and #
# limitations under the License. #
# ------------------------------------------------------------------------ #
import torch
[docs]
def impulse(order, **kwargs):
"""Generate impulse sequence.
See `impulse <https://sp-nitech.github.io/sptk/latest/main/impulse.html>`_
for details.
Parameters
----------
order : int >= 0
Order of sequence, :math:`M`.
**kwargs : additional keyword arguments
See `torch.eye <https://pytorch.org/docs/stable/generated/torch.eye.html>`_.
Returns
-------
out : Tensor [shape=(M+1,)]
Impulse sequence.
Examples
--------
>>> x = diffsptk.impulse(4)
>>> x
tensor([1., 0., 0., 0., 0.])
"""
x = torch.eye(n=1, m=order + 1, **kwargs).squeeze(0)
return x
[docs]
def step(order, value=1, **kwargs):
"""Generate step sequence.
See `step <https://sp-nitech.github.io/sptk/latest/main/step.html>`_
for details.
Parameters
----------
order : int >= 0
Order of sequence, :math:`M`.
value : float
Step value.
**kwargs : additional keyword arguments
See `torch.full <https://pytorch.org/docs/stable/generated/torch.full.html>`_.
Returns
-------
out : Tensor [shape=(M+1,)]
Step sequence.
Examples
--------
>>> x = diffsptk.step(4, 2)
>>> x
tensor([2., 2., 2., 2., 2.])
"""
x = torch.full((order + 1,), float(value), **kwargs)
return x
[docs]
def ramp(arg, end=None, step=1, eps=1e-8, **kwargs):
"""Generate ramp sequence.
See `ramp <https://sp-nitech.github.io/sptk/latest/main/ramp.html>`_
for details.
Parameters
----------
arg : float
If `end` is `None` end value otherwise start value.
end : float
End value.
step : float != 0
Slope.
eps : float
A correction value.
**kwargs : additional keyword arguments
See `torch.arange
<https://pytorch.org/docs/stable/generated/torch.arange.html>`_.
Returns
-------
out : Tensor [shape=(?,)]
Ramp sequence.
Examples
--------
>>> x = diffsptk.ramp(4)
>>> x
tensor([0., 1., 2., 3., 4.])
"""
if end is None:
start = 0
end = arg
else:
start = arg
if step > 0:
end += eps
else:
end -= eps
x = torch.arange(start, end, step, **kwargs)
return x
[docs]
def sin(order, period=None, magnitude=1, **kwargs):
"""Generate sinusoidal sequence.
See `sin <https://sp-nitech.github.io/sptk/latest/main/sin.html>`_
for details.
Parameters
----------
order : int >= 0
Order of sequence, :math:`M`.
period : float > 0
Period.
magnitude : float
Magnitude.
**kwargs : additional keyword arguments
See `torch.arange
<https://pytorch.org/docs/stable/generated/torch.arange.html>`_.
Returns
-------
out : Tensor [shape=(M+1,)]
Sinusoidal sequence.
Examples
--------
>>> x = diffsptk.sin(4)
>>> x
tensor([ 0.0000, 0.9511, 0.5878, -0.5878, -0.9511])
"""
if period is None:
period = order + 1
x = torch.arange(order + 1, **kwargs)
x = torch.sin(x * (2 * torch.pi / period)) * magnitude
return x
[docs]
def train(order, frame_period, norm="power", **kwargs):
"""Generate pulse sequence.
See `train <https://sp-nitech.github.io/sptk/latest/main/train.html>`_
for details.
Parameters
----------
order : int >= 0
Order of sequence, :math:`M`.
frame_period : float >= 1
Frame period.
norm : ['none', 'power', 'magnitude']
Normalization type.
**kwargs : additional keyword arguments
See `torch.zeros <https://pytorch.org/docs/stable/generated/torch.zeros.html>`_.
Returns
-------
out : Tensor [shape=(M+1,)]
Pulse sequence.
Examples
--------
>>> x = diffsptk.train(5, 2.3)
>>> x
tensor([1.5166, 0.0000, 0.0000, 1.5166, 0.0000, 1.5166])
"""
assert 1 <= frame_period
if norm == 0 or norm == "none":
pulse = 1
elif norm == 1 or norm == "power":
pulse = frame_period**0.5
elif norm == 2 or norm == "magnitude":
pulse = frame_period
else:
raise ValueError(f"norm {norm} is not supported.")
frequency = 1 / frame_period
v = torch.full((order + 2,), frequency)
v[0] *= -1
v = torch.floor(torch.cumsum(v, dim=0))
index = torch.ge(torch.diff(v), 1)
x = torch.zeros(order + 1, **kwargs)
x[index] = pulse
return x
[docs]
def nrand(*order, mean=0, stdv=1, var=None, **kwargs):
"""Generate random number sequence.
See `nrand <https://sp-nitech.github.io/sptk/latest/main/nrand.html>`_
for details.
Parameters
----------
order : int >= 0
Order of sequence, :math:`M`.
mean : float
Mean.
stdv : float >= 0
Standard deviation.
var : float >= 0
Variance.
**kwargs : additional keyword arguments
See `torch.randn <https://pytorch.org/docs/stable/generated/torch.randn.html>`_.
Returns
-------
out : Tensor [shape=(M+1,)]
Random value sequence.
Examples
--------
>>> x = diffsptk.nrand(4)
>>> x
tensor([-0.8603, 0.6743, -0.9178, 1.5382, -0.2574])
>>> x = diffsptk.nrand(2, 4)
>>> x
tensor([[-0.2385, -0.0778, -0.0418, -1.6217, 0.1560],
[ 1.6646, 0.8429, 0.9357, -0.5123, 0.9571]])
"""
if var is not None:
stdv = var**0.5
assert 0 <= stdv
if any(isinstance(item, (list, tuple)) for item in order):
order = list(*order)
else:
order = list(order)
order[-1] += 1
x = torch.randn(*order, **kwargs)
x = x * stdv + mean
return x