Audio collecting website for building a speech dataset of 20 isolated words for Digital Signal Processing Spring 2019 course project
The speech dataset collected by this website is on https://github.com/zhongyuchen/DSPSpeech-20.
- speech-classification: CNN and VGG speech classification with interactive website for testing
Each individual is required to record the following 20 words 20 times each word.
| 00 | 01 | 02 | 03 | 04 | 05 | 06 | 07 | 08 | 09 |
|---|---|---|---|---|---|---|---|---|---|
| 数字 | 语音 | 语言 | 识别 | 中国 | 总工 | 北京 | 背景 | 上海 | 商行 |
| 10 | 11 | 12 | 13 | 14 | 15 | 16 | 17 | 18 | 19 |
|---|---|---|---|---|---|---|---|---|---|
| 复旦 | 饭店 | Speech | Speaker | Signal | Process | Open | Close | Project |
| Format | AudioFormat | NumChannels | SampleRate | BitRate | BitDepth | length |
|---|---|---|---|---|---|---|
.wav |
PCM | 1 | 16 kHz | 256 kbps | 16 bits | ~2 s |
These are the important parameters regarding the audio files. If you'd like to get more information about the audio files, you can look closely at the header of the files. While doing so, keep in mind that it's little-endian format!
https://zhongyuchen.github.io/audio-collector/
- wave curve shows the wave pattern of your last recording
- Bezier curve visually shows the volume of what the mic captures
- volume bar shows the volume with its value (0 ~ 100) and a green bar
- Student ID input field input your ID here
- No. input field input a
number(1 ~ 400) here - round (1 ~ 20)
round = number / 20 + 1 - label (00 ~ 19)
label = number % 20 - prompt word will change as the label changes
- mic button starts recording
- ear button plays back the recording
- download button downloads the recording
- It works on Google Chrome and 360 Secure Browser.
- It does NOT work on Safari, Microsoft Edge and Internet Explorer!
- Visit the above website and enter the recording page
- Input your student ID and the
number - Click mic button to start recording and say the prompt word clearly in 2 seconds
- Click ear button to confirm your recording
- Click download button to download your recording
- Repeat the above process as the
numberchanges from 1 to 400 - Put the 20 x 20 correctly named recordings in a folder named after your student ID
- It works best with full screen as the charts are not adaptive!
- If the browser is not compatible, you will be alerted and please use another browser!
- If access to your mic is blocked, the start button will remain disabled and you will also be alerted!
- If you click download button with no ID input, download will fail and you will also be alerted!
- Remember to enable the mic as well!
Zhongyu Chen