88 Hours – Mandarin Speech Data in Noisy Environment by Mobile Phone
- 203 Chinese speakers
- 88 hours
- covering major dialect regions
Datatang has passed the certification of ISO27001 Information Security Management System and ISO9001 Quality Management System.
Spoken Mandarin audio data under noisy environment captured by mobile phone, it is recorded by 203 speakers from all over China, covering all major dialect regions; and a variety of noise scenes such as subways, supermarkets, restaurants, etc., more suitable for real application scenes; it can be used for automatic speech recognition, machine translation, and voiceprint recognition.
- 16kHz, 16bit, uncompressed wav, mono channel
- Recording environment
- noisy, including subway, market, restaurant, street, airport, etc.
- Recording content
- common sentences; letters
- 203 people, 57% of which are male
- Android mobile phone; iPhone
- mandarin (without heavy local accent)
- Transcription content
- text, noise symbols
- Accuracy rate
- 95% (the accuracy rate of noise symbols is not included)
- Application scenarios
- speech recognition, voiceprint recognition