soundsource.py

# @version: 1.0  date: 05/06/2015 by Sidney Barthe
# @author: robin.scheibler@epfl.ch, ivan.dokmanic@epfl.ch, sidney.barthe@epfl.ch
# @copyright: EPFL-IC-LCAV 2015
from __future__ import division, print_function

import numpy as np
from .parameters import constants
from .directivities import Directivity


class SoundSource(object):
    """
    A class to represent sound sources.

    This object represents a sound source in a room by a list containing the original source position
    as well as all the image sources, up to some maximum order.

    It also keeps track of the sequence of generated images and the index of the walls (in the original room)
    that generated the reflection.
    """

    def __init__(
        self,
        position,
        images=None,  # source position
        damping=None,
        generators=None,  # parent source
        walls=None,  # generating wall
        orders=None,
        signal=None,
        delay=0,
        directivity=None,
    ):

        position = np.array(position)
        self.dim = position.shape[0]

        # Check the shape of the passed array
        if self.dim != 2 and self.dim != 3:
            dim_mismatch = True
        else:
            dim_mismatch = False

        if position.ndim == 2 and position.shape[1] == 1:
            position = position[:, 0]

        if position.ndim != 1 or dim_mismatch:
            raise ValueError(
                "The source location of microphones should be provided as an object "
                "that can be converted to a numpy.ndarray. The array should be of "
                "shape `(2,)`, `(2, 1)`, `(3,)`, or `(3, 1)`."
            )

        self.position = position

        if images is None:
            # set to empty list if nothing provided
            self.images = np.asfortranarray(np.array([position], dtype=np.float32).T)
            self.damping = np.array([[1.0]])
            self.generators = np.array([-1], dtype=np.int32)
            self.walls = np.array([-1], dtype=np.int32)
            self.orders = np.array([0], dtype=np.int32)

        else:
            # we need to have damping factors for every image
            if damping is None:
                # set to one if not set
                damping = np.ones((1, images.shape[1]))

            if images.shape[1] != damping.shape[1]:
                raise NameError("Images and damping must have same shape")

            if generators is not None and generators.shape[0] != images.shape[1]:
                raise NameError("Images and generators must have same shape")

            if walls is not None and walls.shape[0] != images.shape[1]:
                raise NameError("Images and walls must have same shape")

            if orders is not None and orders.shape[0] != images.shape[1]:
                raise NameError("Images and orders must have same shape")

            self.images = np.array(images, order="F", dtype=np.float32)
            self.damping = damping
            self.walls = np.array(walls, dtype=np.int32)
            self.generators = np.array(generators, dtype=np.int32)
            self.orders = np.array(orders, dtype=np.int32)

        # store the natural ordering for the images
        self.I = np.arange(self.images.shape[1])

        # the natural ordering is per generation
        self.ordering = "order"

        # The sound signal of the source
        self.signal = signal
        self.delay = delay
        self.max_order = np.max(self.orders)

        # The directivity of the source
        self.directivity = None
        if directivity is not None:
            self.set_directivity(directivity)

    def set_directivity(self, directivity):
        """
        Sets self.directivity as a list of directivities with 1 entry
        """
        assert isinstance(directivity, Directivity)
        self.directivity = directivity

    def add_signal(self, signal):
        """Sets `signal` attribute

        Parameters
        ----------
        signal: ndarray
            a N-length ndarray, representing a sequence of samples generated by the source.
        """
        self.signal = signal

    def distance(self, ref_point):

        return np.sqrt(np.sum((self.images - ref_point[:, np.newaxis]) ** 2, axis=0))

    def set_ordering(self, ordering, ref_point=None):
        """
        Set the order in which we retrieve images sources.
        Can be: 'nearest', 'strongest', 'order'
        Optional argument: ref_point
        """

        self.ordering = ordering

        if ref_point is not None and ref_point.ndim > 1:
            ref_point = ref_point[:, 0]

        if ordering == "nearest":

            if ref_point is None:
                raise NameError("For nearest ordering, a reference point is needed.")

            self.I = self.distance(ref_point).argsort()

        elif ordering == "strongest":

            if ref_point is None:
                raise NameError("For strongest ordering, a reference point is needed.")

            strength = np.linalg.norm(self.damping, 0) / (self.distance(ref_point))
            self.I = strength.argsort()

        elif ordering == "order":

            self.ordering = "order"

        else:
            raise NameError("Ordering can be nearest, strongest, order.")

    def __getitem__(self, index):
        """Overload the bracket operator to access a subset image sources"""

        if isinstance(index, slice) or isinstance(index, int):
            if self.ordering == "order":
                p_orders = np.arange(0, self.max_order + 1)[index]
                # we use the any operator and broadcasting to get match on
                # all image source of order contained in p_orders
                I = np.any(
                    self.orders[:, np.newaxis] == p_orders[np.newaxis, :], axis=1
                )
                s = SoundSource(
                    self.position,
                    images=self.images[:, I],
                    damping=self.damping[:, I],
                    orders=self.orders[I],
                    signal=self.signal,
                    delay=self.delay,
                    generators=self.generators[I],
                    walls=self.walls[I],
                )
            else:
                s = SoundSource(
                    self.position,
                    images=self.images[:, self.I[index]],
                    damping=self.damping[:, self.I[index]],
                    orders=self.orders[self.I[index]],
                    signal=self.signal,
                    delay=self.delay,
                    generators=self.generators[self.I[index]],
                    walls=self.walls[self.I[index]],
                )
        else:
            s = SoundSource(
                self.position,
                images=self.images[:, index],
                damping=self.damping[:, index],
                orders=self.orders[index],
                signal=self.signal,
                delay=self.delay,
                generators=self.generators[index],
                walls=self.walls[index],
            )

        return s

    def get_images(
        self, max_order=None, max_distance=None, n_nearest=None, ref_point=None
    ):
        """
        Keep this for compatibility
        Now replaced by the bracket operator and the setOrdering function.
        """

        # TO DO: Add also n_strongest

        # TO DO: Make some of these thing exclusive (e.g. can't have n_nearest
        # AND n_strongest (although could have max_order AND n_nearest)

        # TO DO: Make this more efficient if bottleneck (unlikely)

        if max_order is None:
            max_order = np.max(self.orders)

        # stack source and all images
        I_ord = self.orders <= max_order
        img = self.images[:, I_ord]

        if n_nearest is not None:
            dist = np.sum((img - ref_point) ** 2, axis=0)
            I_near = dist.argsort()[0:n_nearest]
            img = img[:, I_near]

        return img

    def get_damping(self, max_order=None):
        if max_order is None:
            max_order = len(np.max(self.orders))

        return self.damping[:, self.orders <= max_order]

    def get_rir(self, mic, visibility, Fs, t0=0.0, t_max=None):
        """
        Compute the room impulse response between the source
        and the microphone whose position is given as an
        argument.

        Parameters
        ----------
        mic: ndarray
            microphone position
        visibility: int32
            1 if mic visibile from source, 0 else. Exact type is important for C extension
        Fs: int
            sampling frequency
        t0: float
            time offset, defaults to 0
        t_max: None
            max time, defaults to 1.05 times the propagation time from mic to source
        """

        # fractional delay length
        fdl = constants.get("frac_delay_length")
        fdl2 = (fdl - 1) // 2

        # compute the distance
        dist = self.distance(mic)
        time = dist / constants.get("c") + t0
        if self.damping.shape[0] == 1:
            alpha = self.damping[0, :] / (4.0 * np.pi * dist)
        else:
            raise NotImplementedError("Not implemented for multiple frequency bands")

        # the number of samples needed
        if t_max is None:
            # we give a little bit of time to the sinc to decay anyway
            N = np.ceil((1.05 * time.max() - t0) * Fs)
        else:
            N = np.ceil((t_max - t0) * Fs)

        N += fdl

        t = np.arange(N) / float(Fs)
        ir = np.zeros(t.shape)

        try:
            # Try to use the Cython extension
            from .build_rir import fast_rir_builder

            fast_rir_builder(ir, time, alpha, visibility.astype(np.int32), Fs, fdl)

        except ImportError:
            print("Cython-extension build_rir unavailable. Falling back to pure python")
            # fallback to pure Python implemenation
            from .utilities import fractional_delay

            for i in range(time.shape[0]):
                if visibility[i] == 1:
                    time_ip = int(np.round(Fs * time[i]))
                    time_fp = (Fs * time[i]) - time_ip
                    ir[time_ip - fdl2 : time_ip + fdl2 + 1] += alpha[
                        i
                    ] * fractional_delay(time_fp)

        return ir

    def wall_sequence(self, i):
        """
        Print the wall sequence for the image source indexed by i
        """
        p = self.generators[i]
        if np.isnan(p):
            return []
        w = [self.walls[i]]
        while not np.isnan(p):
            w.append(self.walls[p])
            p = self.generators[p]
        return w


def build_rir_matrix(mics, sources, Lg, Fs, epsilon=5e-3, unit_damping=False):
    """
    A function to build the channel matrix for many sources and microphones

    Parameters
    ----------
    mics: ndarray
        a dim-by-M ndarray where each column is the position of a microphone
    sources: list of pyroomacoustics.SoundSource
        list of sound sources for which we want to build the matrix
    Lg: int
        the length of the beamforming filters
    Fs: int
        the sampling frequency
    epsilon: float, optional
        minimum decay of the sinc before truncation. Defaults to epsilon=5e-3
    unit_damping: bool, optional
        determines if the wall damping parameters are used or not. Default to false.

    Returns
    -------
    the function returns the RIR matrix H =

    ::

         --------------------
         | H_{11} H_{12} ...
         | ...
         |
         --------------------

    where H_{ij} is channel matrix between microphone i and source j.
    H is of type (M*Lg)x((Lg+Lh-1)*S) where Lh is the channel length (determined by epsilon),
    and M, S are the number of microphones, sources, respectively.
    """

    from .beamforming import distance
    from .utilities import low_pass_dirac, convmtx

    for s in sources:
        if s.damping.shape[0] > 1:
            raise NotImplementedError("Multiple frequency bands not supported yet")

    # set the boundaries of RIR filter for given epsilon
    d_min = np.inf
    d_max = 0.0
    dmp_max = 0.0
    for s in range(len(sources)):
        dist_mat = distance(mics, sources[s].images)
        if unit_damping is True:
            dmp_max = np.maximum((1.0 / (4 * np.pi * dist_mat)).max(), dmp_max)
        else:
            dmp_max = np.maximum(
                (sources[s].damping[0, np.newaxis, :] / (4 * np.pi * dist_mat)).max(),
                dmp_max,
            )
        d_min = np.minimum(dist_mat.min(), d_min)
        d_max = np.maximum(dist_mat.max(), d_max)

    t_max = d_max / constants.get("c")
    t_min = d_min / constants.get("c")

    offset = dmp_max / (np.pi * Fs * epsilon)

    # RIR length
    Lh = int((t_max - t_min + 2 * offset) * float(Fs))

    # build the channel matrix
    L = Lg + Lh - 1
    H = np.zeros((Lg * mics.shape[1], len(sources) * L))

    for s in range(len(sources)):
        for r in np.arange(mics.shape[1]):

            dist = sources[s].distance(mics[:, r])
            time = dist / constants.get("c") - t_min + offset
            if unit_damping == True:
                dmp = 1.0 / (4 * np.pi * dist)
            else:
                dmp = sources[s].damping[0, :] / (4 * np.pi * dist)

            h = low_pass_dirac(time[:, np.newaxis], dmp[:, np.newaxis], Fs, Lh).sum(
                axis=0
            )
            H[r * Lg : (r + 1) * Lg, s * L : (s + 1) * L] = convmtx(h, Lg).T

    return H