Add: Doc & tools about coherence

This commit is contained in:
Peter.Yang 2018-03-12 06:39:43 +00:00
parent a98c728f33
commit 829ae64abe
2 changed files with 62 additions and 2 deletions

View file

@ -97,7 +97,7 @@ If you want to change the alsa settings, You can use `sudo alsactl --file=/etc/v
## 6-Mics Circular Array Kit
[![](https://user-images.githubusercontent.com/3901856/37193936-d727868a-23a6-11e8-8096-2c531ff479fc.png)]()
[![](https://user-images.githubusercontent.com/3901856/37268348-6adef768-2600-11e8-8861-588b1c3ea142.png)]()
The 6 Mics Circular Array Kit uses ac108 x 2 / ac101 x 1 / micphones x 6, includes 8 ADCs and 2 DACs.
@ -195,7 +195,7 @@ plughw:CARD=seeed8micvoicec,DEV=0
In contrast to 6-Mics Circular Array Kit for Raspberry Pi,
the difference is only first 4 input channels are valid capturing data.
#### Test:
### Usage:
```bash
#for ReSpeaker 2-mic
#It will capture sound an playback on hw:1
@ -221,6 +221,22 @@ aplay -D ac101 a.wav
**Note: for developer using 6-Mics Circular Array Kit(or 4-Mics Linear Array Kit) doing capturing & playback the same time,
capturing must be start first, or else the capturing channels will miss order.**
### Coherence
Estimate the magnitude squared coherence using Welchs method.
![4-mics-linear-array-kit coherence](https://user-images.githubusercontent.com/3901856/37268469-2869dd34-2601-11e8-9a4e-a831b3af82fe.png)
Note: 'CO 1-2' means the coherence between channel 1 and channel 2.
```bash
# How to get the coherence of the captured audio(a.wav for example).
sudo apt install python-numpy python-scipy python-matplotlib
python tools/coherence.py a.wav
Requirement of the input audio file:
- formt: WAV(Microsoft) signed 16-bit PCM
- channels: >=2
```
### uninstall seeed-voicecard
If you want to upgrade the driver , you need uninstall the driver first.

44
tools/coherence.py Normal file
View file

@ -0,0 +1,44 @@
"""
Estimate the magnitude squared coherence estimate,
- requirements
sudo apt install python-numpy python-scipy python-matplotlib
"""
import sys
import wave
import numpy as np
from scipy import signal
import matplotlib.pyplot as plt
if len(sys.argv) < 2:
print('Usage: python {} audio.wav'.format(sys.argv[0]))
sys.exit(1)
wav = wave.open(sys.argv[1], 'rb')
channels = wav.getnchannels()
frames = wav.readframes(wav.getnframes())
fs = wav.getframerate()
wav.close()
print("channels: %d" % channels)
print("rate : %d" % fs)
print("frames : %d" % wav.getnframes())
array = np.fromstring(frames, dtype='int16')
ch0 = array[0::channels]
fig, ax = plt.subplots()
for ch in range(1, channels):
f, c = signal.coherence(ch0, array[ch::channels], fs, nperseg=1024)
ax.semilogy(f, c, label="CO 1-%d" % (ch + 1))
legend = ax.legend(loc='lower right', shadow=True, fontsize='small')
plt.xlabel('frequency [Hz]')
plt.ylabel('Coherence')
plt.show()