lescidium commited on
Commit
4d1d982
·
1 Parent(s): f1c103f

Add example usage

Browse files

Slightly change readme

Files changed (2) hide show
  1. README.md +17 -5
  2. personal_mic_wer.png +3 -0
README.md CHANGED
@@ -9,6 +9,7 @@ pretty_name: Audio Device
9
  # Parallel Recordings
10
  This dataset is a small corpus of audio recorded in parallel.
11
  This data can be used to quantify the difference in recording quality between different audio devices.
 
12
 
13
  ## Organization
14
  There are two separate experiments.
@@ -16,14 +17,13 @@ There are two separate experiments.
16
  2) Speakerphone
17
 
18
  The original data for both experiments are located in the`full_length_audio` directory.
19
-
20
- ## Methodology
21
  Each experiment was performed by recording simultaneously on multiple devices.
22
  Example sentences were prepared beforehand and read during the simultaneous recording.
 
23
 
24
  ### Personal Microphone
25
  **Devices:**
26
- * AT8: Audio Technica AT8
27
  * ATR: Audio Technica ATR4800-USB
28
  * bltin: Built-In microphone on unknown laptop
29
  * boya: Boya BY-W4, Ultracompact 2.4GHz Four-channel Wireless Microphone System
@@ -34,12 +34,24 @@ Each device was pointed in the same direction toward the speaker. The speaker re
34
  * Indexes 30-39: Distance = 1 yard (914.4cm)
35
  All other indexes correspond to user speaking to themselves at varying and inconsistent distances.
36
 
 
 
 
 
 
 
 
 
 
 
 
 
37
  ### Speakerphones
38
  **Devices:**
39
  * respeaker: SeeedStudio ReSpeaker USB Mic Array
40
- * jabra: Jabra Speakerphone
41
  * sanwa: Sanwa Supply 400-MC011
42
- * yamaha: Yamaha Speakerphone
43
  * ntt: NTT Advanced Technology FR-1100
44
 
45
  Each device was placed in the same location in the center of the room. The speaker read sentences five at a time, moving position each time. The positions were roughly assigned and can be best described as close-rear-left, far-front-left, far-front-right, close-rear-right (in order). The final paragraph was read in the same location as the last 5 sentences (close-rear-right).
 
9
  # Parallel Recordings
10
  This dataset is a small corpus of audio recorded in parallel.
11
  This data can be used to quantify the difference in recording quality between different audio devices.
12
+ An example of this quantification using WER is [shown below](#example-of-usage).
13
 
14
  ## Organization
15
  There are two separate experiments.
 
17
  2) Speakerphone
18
 
19
  The original data for both experiments are located in the`full_length_audio` directory.
 
 
20
  Each experiment was performed by recording simultaneously on multiple devices.
21
  Example sentences were prepared beforehand and read during the simultaneous recording.
22
+ The recordings were cut into bits that could fit within the context of most modern ASR models (ie. < 30s).
23
 
24
  ### Personal Microphone
25
  **Devices:**
26
+ * AT8: Audio Technica Digital 2.4GHz System 10 + ATW-T1007 + AT857SL
27
  * ATR: Audio Technica ATR4800-USB
28
  * bltin: Built-In microphone on unknown laptop
29
  * boya: Boya BY-W4, Ultracompact 2.4GHz Four-channel Wireless Microphone System
 
34
  * Indexes 30-39: Distance = 1 yard (914.4cm)
35
  All other indexes correspond to user speaking to themselves at varying and inconsistent distances.
36
 
37
+ #### Example of Usage
38
+ This data was used to measure microphone quality in the context of ASR. The metric used was WER from the output of `WhisperModel` from `faster_whisper`. The follow parameters were used:
39
+ ```
40
+ whisper = WhisperModel(
41
+ model_size_or_path="deepdml/faster-whisper-large-v3-turbo-ct2",
42
+ device="cuda",
43
+ compute_type="float16",
44
+ )
45
+ ```
46
+
47
+ ![Image of graph showing experiment results](personal_mic_wer.png)
48
+
49
  ### Speakerphones
50
  **Devices:**
51
  * respeaker: SeeedStudio ReSpeaker USB Mic Array
52
+ * jabra: Jabra Speak2 75, PHS060Wa
53
  * sanwa: Sanwa Supply 400-MC011
54
+ * yamaha: Yamaha Unified Communications Speakerphone YVC-330
55
  * ntt: NTT Advanced Technology FR-1100
56
 
57
  Each device was placed in the same location in the center of the room. The speaker read sentences five at a time, moving position each time. The positions were roughly assigned and can be best described as close-rear-left, far-front-left, far-front-right, close-rear-right (in order). The final paragraph was read in the same location as the last 5 sentences (close-rear-right).
personal_mic_wer.png ADDED

Git LFS Details

  • SHA256: 02b38573ab0dc791b474667ef8fa79e8ccfaa03557af896d58327f47636a5588
  • Pointer size: 131 Bytes
  • Size of remote file: 556 kB