Merge pull request #666 from danibene/feature/signal_interpolate_nans…

…_extrapolate

[Feature] Modify signal_interpolate for input with NaNs and extrapolation

NEWS.rst

@@ -2,6 +2,32 @@ News
 =====
 
 
+
+
+
+0.2.1
+-------------------
+New Features
++++++++++++++
+
+* Allow for input with NaNs and extrapolation in `signal_interpolate()`
+* Add argument `method` in `find_outliers()`
+
+
+
+
+
+0.2.0
+-------------------
+New Features
++++++++++++++
+
+* Add new time-domain measures in `hrv_time()`: `Prc20NN`, `Prc80NN`, `MinNN`, and `MaxNN`
+
+
+
+
+
 0.1.6
 -------------------
 

neurokit2/signal/signal_interpolate.py

@@ -3,17 +3,19 @@
 import scipy.interpolate
 
 
-def signal_interpolate(x_values, y_values, x_new=None, method="quadratic"):
+def signal_interpolate(
+    x_values, y_values=None, x_new=None, method="quadratic", fill_value=None
+):
     """**Interpolate a signal**
-
     Interpolate a signal using different methods.
-
     Parameters
     ----------
     x_values : Union[list, np.array, pd.Series]
         The samples corresponding to the values to be interpolated.
     y_values : Union[list, np.array, pd.Series]
-        The values to be interpolated.
+        The values to be interpolated. If not provided, any NaNs in the x_values
+        will be interpolated with :func:`_signal_interpolate_nan`,
+        considering the x_values as equally spaced.
     x_new : Union[list, np.array, pd.Series] or int
         The samples at which to interpolate the y_values. Samples before the first value in x_values
         or after the last value in x_values will be extrapolated. If an integer is passed, nex_x
@@ -29,27 +31,28 @@ def signal_interpolate(x_values, y_values, x_new=None, method="quadratic"):
         order of the spline interpolator to use.
         See `here <https://docs.scipy.org/doc/scipy/reference/generated/scipy.interpolate.
         PchipInterpolator.html>`_ for details on the ``"monotone_cubic"`` method.
-
+    fill_value : float or tuple or str
+        If a ndarray (or float), this value will be used to fill in for
+        requested points outside of the data range.
+        If a two-element tuple, then the first element is used as a fill value
+        for x_new < x[0] and the second element is used for x_new > x[-1].
+        If "extrapolate", then points outside the data range will be extrapolated.
+        If not provided, then the default is ([y_values[0]], [y_values[-1]]).
     Returns
     -------
     array
         Vector of interpolated samples.
-
     Examples
     --------
     .. ipython:: python
-
       import numpy as np
       import neurokit2 as nk
       import matplotlib.pyplot as plt
-
       # Generate Simulated Signal
       signal = nk.signal_simulate(duration=2, sampling_rate=10)
-
       # We want to interpolate to 2000 samples
       x_values = np.linspace(0, 2000, num=len(signal), endpoint=False)
       x_new = np.linspace(0, 2000, num=2000, endpoint=False)
-
       # Visualize all interpolation methods
       @savefig p_signal_interpolate1.png scale=100%
       nk.signal_plot([
@@ -65,42 +68,84 @@ def signal_interpolate(x_values, y_values, x_new=None, method="quadratic"):
       plt.scatter(x_values, signal, label="original datapoints", zorder=3)
       @suppress
       plt.close()
-
     """
     # Sanity checks
+    if x_values is None:
+        raise ValueError(
+            "NeuroKit error: signal_interpolate(): x_values must be provided."
+        )
+    elif y_values is None:
+        # for interpolating NaNs
+        return _signal_interpolate_nan(x_values, method=method, fill_value=fill_value)
+
     if len(x_values) != len(y_values):
         raise ValueError("x_values and y_values must be of the same length.")
 
     if isinstance(x_new, int):
         if len(x_values) == x_new:
             return y_values
     else:
-        if len(x_values) == len(x_new):
+      # if x_values is identical to x_new, no need for interpolation
+        if np.all(x_values == x_new):
             return y_values
-
     if method == "monotone_cubic":
         interpolation_function = scipy.interpolate.PchipInterpolator(
             x_values, y_values, extrapolate=True
         )
     else:
+        if fill_value is None:
+            fill_value = ([y_values[0]], [y_values[-1]])
         interpolation_function = scipy.interpolate.interp1d(
             x_values,
             y_values,
             kind=method,
             bounds_error=False,
-            fill_value=([y_values[0]], [y_values[-1]]),
+            fill_value=fill_value,
         )
-
     if isinstance(x_new, int):
         x_new = np.linspace(x_values[0], x_values[-1], x_new)
-
     interpolated = interpolation_function(x_new)
 
-    if method == "monotone_cubic":
-        # Swap out the cubic extrapolation of out-of-bounds segments generated by
-        # scipy.interpolate.PchipInterpolator for constant extrapolation akin to the behavior of
-        # scipy.interpolate.interp1d with fill_value=([y_values[0]], [y_values[-1]].
-        interpolated[: int(x_values[0])] = interpolated[int(x_values[0])]
-        interpolated[int(x_values[-1]) :] = interpolated[int(x_values[-1])]
-
+    if method == "monotone_cubic" and fill_value != "extrapolate":
+        # Find the index of the new x value that is closest to the first original x value
+        first_index = np.argmin(np.abs(x_new - x_values[0]))
+        # Find the index of the new x value that is closest to the last original x value
+        last_index = np.argmin(np.abs(x_new - x_values[-1]))
+
+        if fill_value is None:
+            # Swap out the cubic extrapolation of out-of-bounds segments generated by
+            # scipy.interpolate.PchipInterpolator for constant extrapolation akin to the behavior of
+            # scipy.interpolate.interp1d with fill_value=([y_values[0]], [y_values[-1]].
+            fill_value = ([interpolated[first_index]], [interpolated[last_index]])
+        elif isinstance(fill_value, float) or isinstance(fill_value, int):
+            # if only a single integer or float is provided as a fill value, format as a tuple
+            fill_value = ([fill_value], [fill_value])
+
+        interpolated[: first_index] = fill_value[0]
+        interpolated[last_index + 1:] = fill_value[1]
+
     return interpolated
+
+
+def _signal_interpolate_nan(values, method="quadratic", fill_value=None):
+    if np.any(np.isnan(values)):
+        # assume that values are evenly spaced
+        # x_new corresponds to the indices of all values, including missing
+        x_new = np.arange(len(values))
+        not_missing = np.where(np.invert(np.isnan(values)))[0]
+
+        # remove the missing values
+        y_values = values[not_missing]
+
+        # x_values corresponds to the indices of only non-missing values
+        x_values = x_new[not_missing]
+
+        # interpolate to get the values at the indices where they are missing
+        return signal_interpolate(x_values=x_values,
+                                  y_values=y_values,
+                                  x_new=x_new,
+                                  method=method,
+                                  fill_value=fill_value)
+    else:
+        # if there are no missing values, return original values
+        return values