VB_diarization_demo.py

#!/usr/bin/env python

# Copyright 2013-2017 Lukas Burget (burget@fit.vutbr.cz)
#
# 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.

#
# Revision History
#   L. Burget   16/07/13 01:00AM - original version
#   Quan Wang   07/02/19 11:00AM - refactoring

from __future__ import absolute_import
from __future__ import division
from __future__ import print_function

import numpy as np
import vbdiar

# read UBM and iXtractor (total variability) subspace
m = np.loadtxt("data/UBM_means.txt", dtype=np.float32)
iE = 1.0 / np.loadtxt("data/UBM_vars.txt", dtype=np.float32)
w = np.loadtxt("data/UBM_weights.txt", dtype=np.float32)
V = np.loadtxt("data/iXtractor.txt", dtype=np.float32).reshape(-1, *m.shape)

# load MFCC features
X = np.loadtxt("data/features.txt")

# load reference per-frame labels
# (0: silence; 1: both; 2: speaker; A 3: speaker B)
ref = np.loadtxt("data/ref.txt", dtype=int)

# keep only frames with one speaker (to make scoring simple)
X = X[ref > 1]
ref = ref[ref > 1] - 2

VtiEV = vbdiar.precalculate_VtiEV(V, iE)

q = None
# q = vbdiar.frame_labels2posterior_mx(ref) # initialize from the reference

# runing with one 25 frame (0.25s) resolution; about 2.5 faster
q, sp, L = vbdiar.VB_diarization(
    X, m, iE, w, V, sp=None, q=None, maxSpeakers=2, maxIters=10, VtiEV=VtiEV,
    downsample=25, alphaQInit=100.0, sparsityThr=0.001, epsilon=1e-6, minDur=1,
    loopProb=0.9, statScale=0.3, llScale=1.0, ref=ref, plot=False)

# runing with one frame resolution
q, sp, L = vbdiar.VB_diarization(
    X, m, iE, w, V, sp=None, q=None, maxSpeakers=2, maxIters=10, VtiEV=VtiEV,
    downsample=1, alphaQInit=100.0, sparsityThr=0.001, epsilon=1e-6, minDur=1,
    loopProb=0.998, statScale=0.3, llScale=1.0, ref=ref, plot=True)