minimum_variance_distortioless_response_test.py

# -*- coding: utf-8 -*-
"""
Created on Tue Jan 15 17:49:51 2019

@author: a-kojima
"""
import numpy as np
import soundfile as sf
from beamformer import util
from beamformer import minimum_variance_distortioless_response as mvdr

SAMPLING_FREQUENCY = 16000
FFT_LENGTH = 512
FFT_SHIFT = 256
ENHANCED_WAV_NAME = './output/enhanced_speech_mvdr.wav'
MIC_ANGLE_VECTOR = np.array([0, 60, 120, 180, 270, 330])
LOOK_DIRECTION = 0
MIC_DIAMETER = 0.1

def multi_channel_read(prefix=r'./sample_data/20G_20GO010I_STR.CH{}.wav',
                       channel_index_vector=np.array([1, 2, 3, 4, 5, 6])):
    wav, _ = sf.read(prefix.replace('{}', str(channel_index_vector[0])), dtype='float32')
    wav_multi = np.zeros((len(wav), len(channel_index_vector)), dtype=np.float32)
    wav_multi[:, 0] = wav
    for i in range(1, len(channel_index_vector)):
        wav_multi[:, i] = sf.read(prefix.replace('{}', str(channel_index_vector[i])), dtype='float32')[0]
    return wav_multi

multi_channels_data = multi_channel_read()


complex_spectrum, _ = util.get_3dim_spectrum_from_data(multi_channels_data, FFT_LENGTH, FFT_SHIFT, FFT_LENGTH)

mvdr_beamformer = mvdr.minimum_variance_distortioless_response(MIC_ANGLE_VECTOR, MIC_DIAMETER, sampling_frequency=SAMPLING_FREQUENCY, fft_length=FFT_LENGTH, fft_shift=FFT_SHIFT)

steering_vector = mvdr_beamformer.get_sterring_vector(LOOK_DIRECTION)

spatial_correlation_matrix = mvdr_beamformer.get_spatial_correlation_matrix(multi_channels_data)

beamformer = mvdr_beamformer.get_mvdr_beamformer(steering_vector, spatial_correlation_matrix)

enhanced_speech = mvdr_beamformer.apply_beamformer(beamformer, complex_spectrum)

sf.write(ENHANCED_WAV_NAME, enhanced_speech / np.max(np.abs(enhanced_speech)) * 0.65, SAMPLING_FREQUENCY)