1. The sampling rate (sample rate)
It refers to the number of times the audio signal recording apparatus in a second sample,The higher sampling frequency, the more the number of samples collected per unit time, to obtain the original waveform the waveform of the closer, the better the sound quality.
Currently only supports voice recognition service16000Hzwith8000HzTwo sampling rate , which is generally 8000Hz telephone service use, and the rest use 16000Hz.
2. Sampling digit (sample size)
I.e. sample values or sample values. It is used to measure a parameter fluctuating sound, the sound card can also be said of the resolution.The greater its value, the higher the resolution, the ability of the sound.
Each amplitude sample data record, depending on the size of the sampling accuracy of the sample bits:
- 1 byte (i.e. 8bit) can record number 256, i.e. the amplitude can be divided into 256 gradations;
- 2-byte (i.e. 16bit) may be fine to the number 65536, is CD (clear sound) of the standard;
- 4 bytes (that is, 32bit) can subdivide 4294967296 amplitude levels
3. The speech coding (format)
It refers to the way that the voice data storage and transmission. Voice coding and voice file formats are different . .WAV file format such as a common, in particular in its definition will be encoded voice data header, whereinThe audio data is typically encoded using the PCM, But there may be other AMR encoding.
4. Channel (sound chanel)
Refers to sound recorded position acquired at different spatial independent audio signals, the number of channels is, the number of the sound source when the sound recording.Common audio data is mono or two-channel (stereo)。
Recording test corpus related settings are as follows:
16000Hz sampling rate selection, to select 16 bits per sample (clear sound), selecting PCM coding speech encoding, channel selection mono or binaural
