一、概述
本节我们使用 pyroomacoustics库生成多通道仿真数据，方便对算法的验证。
二、数据仿真
2.1 数据仿真代码
import pyroomacoustics as pra
import numpy as np
import matplotlib.pyplot as plt
import librosa
# 1、创建房间
# 所需的混响时间和房间的尺寸
rt60_tgt = 0.2 # 所需的混响时间，秒
# room_dim = [9, 7.5, 3.5] # 我们定义了一个9m x 7.5m x 3.5m的房间，米
room_dim = [9, 7.5] # 我们定义了一个9m x 7.5m x 3.5m的房间，米
# 我们可以使用Sabine’s公式来计算壁面能量吸收和达到预期混响时间所需的ISM的最大阶数(RT60，即RIR衰减60分贝所需的时间)
e_absorption, max_order = pra.inverse_sabine(rt60_tgt, room_dim) # 返回 墙壁吸收的能量 和 允许的反射次数
# 我们还可以自定义 墙壁材料 和 最大反射次数
# m = pra.Material(energy_absorption="hard_surface") # 定义 墙的材料，我们还可以定义不同墙面的的材料
# max_order = 3
room = pra.ShoeBox(room_dim, fs=16000, materials=pra.Material(e_absorption), max_order=max_order)
#坐标值计算
# angle = 300 * np.pi / 180
# r = 2
#
# x_center = 4.5
# y_center = 3
# z_center = 1.0
#
# xx = r * np.cos(angle) + x_center
# yy = r * np.sin(angle) + y_center
# zz = z_center
# 在房间内创建一个位于[2.5,3.73,1.76]的源，从0.3秒开始向仿真中发出wav文件的内容
audio, _ = librosa.load("input/role1.wav",sr=16000) # 导入一个单通道语音作为源信号 source signal
# room.add_source([6.5, 3.0, 1.0], signal=audio, delay=0.3)
room.add_source([6.0, 3.0], signal=audio, delay=0)
audio2, _ = librosa.load("input/role2.wav",sr=16000) # 导入一个单通道语音作为源信号 source signal
# room.add_source([4.5, 1.0, 1.0], signal=audio2, delay=0.3)
room.add_source([1.5, 3.0], signal=audio2, delay=0)
# 3、在房间放置麦克风
# 定义麦克风的位置：(ndim, nmics) 即每个列包含一个麦克风的坐标
# 在这里我们创建一个带有两个麦克风的数组，
# 分别位于[6.3,4.87,1.2]和[6.3,4.93,1.2]。
# mic_locs = np.c_[
# [4.5463, 3.0, 1.2], # mic 1
# [4.52315, 2.9599, 1.2], # mic 2
# [4.47685, 2.9599, 1.2], # mic 3
# [4.4537, 3.0, 1.2], # mic 4
# [4.47685, 3.0401, 1.2], # mic 5
# [4.52315, 3.0401, 1.2] # mic 6
# ]
mic_locs = np.c_[
[4.5463, 3.0], # mic 1
[4.52315, 2.9599], # mic 2
[4.47685, 2.9599], # mic 3
[4.4537, 3.0], # mic 4
[4.47685, 3.0401], # mic 5
[4.52315, 3.0401] # mic 6
]
room.add_microphone_array(mic_locs) # 最后将麦克风阵列放在房间里
# # define the locations of the microphones
# mic_num = 6
# mic_radius = 0.0463
# R = pra.circular_2D_array([4,4], mic_num, 0, mic_radius)
#
# room.add_microphone_array(pra.MicrophoneArray(R, room.fs))
# 4、创建房间冲击响应（Room Impulse Response）
room.compute_rir()
# 5、模拟声音传播，每个源的信号将与相应的房间脉冲响应进行卷积。卷积的输出将在麦克风上求和。
room.simulate()
# 保存所有的信号到wav文件
room.mic_array.to_wav("./output/simulate_role1_0_t60_0.2_role2_180_t60_0.2.wav", norm=True, bitdepth=np.int16)
# 测量混响时间
rt60 = room.measure_rt60()
print("The desired RT60 was {}".format(rt60_tgt))
print("The measured RT60 is {}".format(rt60[1, 0]))
# plt.figure()
# # 绘制其中一个RIR. both can also be plotted using room.plot_rir()
# rir_1_0 = room.rir[1][0] # 画出 mic 1和 source 0 之间的 RIR
# plt.subplot(2, 1, 1)
# plt.plot(np.arange(len(rir_1_0)) / room.fs, rir_1_0)
# plt.title("The RIR from source 0 to mic 1")
# plt.xlabel("Time [s]")
#
# # 绘制 microphone 1 处接收到的信号
# plt.subplot(2, 1, 2)
# plt.plot(np.arange(len(room.mic_array.signals[1, :])) / room.fs, room.mic_array.signals[1, :])
# plt.title("Microphone 1 signal")
# plt.xlabel("Time [s]")
#
# plt.tight_layout()
# plt.show()
fig, ax = room.plot()
plt.grid()
plt.show()
2.2 仿真结果
（1）相对位置
（2）生成数据
2.3 respeaker core v2 数据录制（环阵半径0.463）
#define ALSA_PCM_NEW_HW_PARAMS_API
#include <alsa/asoundlib.h>
#include <stdio.h>
#include <sys/types.h>
#include <sys/stat.h>
#include <stdlib.h>
int main(int argc, char *argv[]) {
long loops;
int rc;
int size;
unsigned int val;
int dir;
char *buffer;
snd_pcm_t *handle;
snd_pcm_hw_params_t *params;
snd_pcm_uframes_t frames;
/*以录制模式打开*/
/* Open PCM device for recording (capture). */
rc = snd_pcm_open( &handle, "default", SND_PCM_STREAM_CAPTURE, 0);
if (rc < 0) {
fprintf(stderr, "unable to open pcm device");
exit(EXIT_FAILURE);
}
/*分配一个参数对象*/
/* Allocate a hardware parameters object. */
snd_pcm_hw_params_alloca(¶ms);
/*初始化参数对象*/
/* Fill it in with default values. */
rc = snd_pcm_hw_params_any(handle, params);
if (rc < 0) {
printf("Err\n");
}
/* Set the desired hardware parameters. */
/*交错模式*/
/* Interleaved mode */
rc = snd_pcm_hw_params_set_access(handle, params,
SND_PCM_ACCESS_RW_INTERLEAVED);
if (rc < 0) {
printf("Err\n");
}
/*PCM格式*/
/* Signed 16-bit little-endian format */
rc = snd_pcm_hw_params_set_format(handle, params,
SND_PCM_FORMAT_S16_LE);
if (rc < 0) {
printf("Err\n");
}
/*设置通道数*/
/* 8 channels (stereo) */
rc = snd_pcm_hw_params_set_channels(handle, params, 8);
if (rc < 0) {
printf("Err\n");
}
/*设置采样率*/
/* 44100 bits/second sampling rate (CD quality) */
val = 16000;
rc = snd_pcm_hw_params_set_rate_near(handle, params,
&val, &dir);
if (rc < 0) {
printf("Err\n");
}
/*没周期的帧数*/
/* Set period size to 32 frames. */
frames = 32;
rc = snd_pcm_hw_params_set_period_size_near(handle,
params, &frames, &dir);
if (rc < 0) {
printf("Err\n");
}
/* Write the parameters to the driver */
rc = snd_pcm_hw_params(handle, params);
if (rc < 0) {
fprintf(stderr,
"unable to set hw parameters: %s/n",
snd_strerror(rc));
exit(1);
}
/* Use a buffer large enough to hold one period */
rc = snd_pcm_hw_params_get_period_size(params,
&frames, &dir);
if (rc < 0) {
printf("Err\n");
}
size = frames * 2 * 8; /* 2 bytes/sample, 8 channels */
buffer = (char *) malloc(size);
/* We want to loop for 5 seconds */
rc = snd_pcm_hw_params_get_period_time(params, &val, &dir);
loops = 5000000 / val;
//printf("====================:%d\n", frames);
//printf("====================\n");
printf("++++ start record!\n");
FILE *fp = fopen("record_16bit_8channal_16k.pcm", "wb");
while (loops > 0) {
//loops--;
rc = snd_pcm_readi(handle, buffer, frames);
if (rc == -EPIPE) {
/* EPIPE means overrun */
fprintf(stderr, "overrun occurred/n");
//把PCM流置于PREPARED状态，这样下次我们向该PCM流中数据时，它就能重新开始处理数据。
snd_pcm_prepare(handle);
} else if (rc < 0) {
fprintf(stderr,
"error from read: %s/n",
snd_strerror(rc));
} else if (rc != (int)frames) {
fprintf(stderr, "short read, read %d frames/n", rc);
}
short *buf = (short*)buffer;
for (int i = 0; i < frames; i++) {
buf[i] *= 1;
}
rc = fwrite(buffer, 1, size, fp);
if (rc != size) {
fprintf(stderr,
"short write: wrote %d bytes/n", rc);
break;
}
}
//调用snd_pcm_drain把所有挂起没有传输完的声音样本传输完全
rc = snd_pcm_drain(handle);
//关闭该音频流，释放之前动态分配的缓冲区，退出
rc = snd_pcm_close(handle);
free(buffer);
fclose(fp);
return 0;
}
三、总结
本节通过单通道数据，设置房间冲激响应函数参数，模拟麦克风阵列，生成多通道仿真数据，方便后续对算法的仿真。
扫一扫
被折叠的 条评论
为什么被折叠?
到【灌水乐园】发言
