Chromium Code Reviews
chromiumcodereview-hr@appspot.gserviceaccount.com (chromiumcodereview-hr) | Please choose your nickname with Settings | Help | Chromium Project | Gerrit Changes | Sign out
(535)

Side by Side Diff: webrtc/modules/audio_processing/test/py_conversational_speech/README.md

Issue 2752793002: Reland "C++ porting of the initial python script for conversational speech generation." (Closed)
Patch Set: deps fixed Created 3 years, 9 months ago
Use n/p to move between diff chunks; N/P to move between comments. Draft comments are only viewable by you.
Jump to:
View unified diff | Download patch
OLDNEW
(Empty)
1 #Conversational Speech generator tool
2
3 Python tool to generate multiple-end audio tracks to simulate conversational
4 speech with two or more participants.
5
6 The input to the tool is a directory containing a number of audio tracks and
7 a text file indicating how to time the sequence of speech turns (see the Example
8 section).
9
10 Since the timing of the speaking turns is specified by the user, the generated
11 tracks may not be suitable for testing scenarios in which there is unpredictable
12 network delay (e.g., end-to-end RTC assessment).
13
14 Instead, the generated pairs can be used when the delay is constant (obviously
15 including the case in which there is no delay).
16 For instance, echo cancellation in the APM module can be evaluated using two-end
17 audio tracks as input and reverse input.
18
19 By indicating negative and positive time offsets, one can reproduce cross-talk
20 and silence in the conversation.
21
22 IMPORTANT: **the whole code has not been landed yet.**
23
24 ###Example
25
26 For each end, there is a set of audio tracks, e.g., a1, a2 and a3 (speaker A)
27 and b1, b2 (speaker B).
28 The text file with the timing information may look like this:
29
30 ```
31 A a1 0
32 B b1 0
33 A a2 100
34 B b2 -200
35 A a3 0
36 A a4 0
37 ```
38
39 The first column indicates the speaker name, the second contains the audio track
40 file names, and the third the offsets (in milliseconds) used to concatenate the
41 chunks.
42
43 Assume that all the audio tracks in the example above are 1000 ms long.
44 The tool will then generate two tracks (A and B) that look like this:
45
46 **Track A**
47 ```
48 a1 (1000 ms)
49 silence (1100 ms)
50 a2 (1000 ms)
51 silence (800 ms)
52 a3 (1000 ms)
53 a4 (1000 ms)
54 ```
55
56 **Track B**
57 ```
58 silence (1000 ms)
59 b1 (1000 ms)
60 silence (900 ms)
61 b2 (1000 ms)
62 silence (2000 ms)
63 ```
64
65 The two tracks can be also visualized as follows (one characheter represents
66 100 ms, "." is silence and "*" is speech).
67
68 ```
69 t: 0 1 2 3 4 5 6 (s)
70 A: **********...........**********........********************
71 B: ..........**********.........**********....................
72 ^ 200 ms cross-talk
73 100 ms silence ^
74 ```
OLDNEW

Powered by Google App Engine
This is Rietveld 408576698