There are several common noise removal methods:

1. Noise door:

Set the threshold value of a level, filter out all the signal levels lower than this threshold, and pass all the signal levels higher than the threshold value (here, the signal level refers to the level of the sum of signal and noise levels). This kind of sound data with high signal-to-noise ratio does not need denoising. Rvox in wave plug-in is the effector of such a noise gate. How many parameters are appropriate? It is generally about - 50 ~ - 40dB, but the size of the background noise is not certain in practice, so you need to try it yourself. The correct position is that you can't hear the set noise, but the small voice announced by the human voice is not filtered out.

2. Sampling denoising method:

This is a method for audio processing software to remove continuous and stable scene noise. The principle of noise removal is to sample the waveform samples of noise, and then analyze the waveform and sampled noise samples of the whole section of data to automatically remove noise. The advantage of this noise removal is that it can completely remove the noise, but the disadvantage is that it can damage the sound quality of the original human voice. The lower the signal-to-noise ratio, the greater the damage. Through this noise removal, the sound has a strong metallic smell. Therefore, this method is generally not recommended to denoise the singing voice. The key to solving the problem is to control the recording to a high signal-to-noise ratio in the early stage as far as possible.

3. Other noise removal methods:

Analyze the frequency band of the noise place with the spectrum, and attenuate the frequency band signal level of the bottom noise place through the adjustment of Eq. However, the signals of the original vocal band are also attenuated, which will pose a new problem, so they are also used less.

EQ treatment:

The spectrum dispersion of human voice source is special. As far as its pronunciation method is concerned, it has three parts: first, the music generated by vocal cord vibration. The pronunciation of this part is flexible, and the spectrum changes greatly with different pitch and pronunciation methods; The second is the low-frequency Kai tone generated by nasal resonance. Frequency equalization can roughly separate these three parts of the spectrum. The frequency band used for nasal sound is below 500Hz, the midpoint frequency of equalization is generally 80-150hz, and the equalization bandwidth is 4 octaves.

For example, 100Hz can be set as the midpoint of frequency equalization, the equalization curve should transition steeply from 100-400hz, and the scale of equalization gain can be + 10dB ~ - 6dB. Here we should be reminded that the monitoring speaker during the adjustment of this project shall not use the box with weak low-frequency pronunciation, so as to avoid the aggravation of nasal sound inadvertently. The spectrum of human voice and dental sound is scattered above 4kHz. Since this frequency band contains part of the music spectrum, it is advocated that the frequency band of tooth sound should be 6 ~ 16KHz, the equalization bandwidth should be 3 octaves, the equalization midpoint frequency is generally 1 / 2 octaves, and the equalization midpoint frequency is 6800hz. The equalization gain is low and can be adjusted down to - 10dB.

From the above analysis, it can be seen that during the frequency equalization of human voice, the frequency band enhancement for an outstanding sound feeling should try to use the steep curve broadband equalization. This is to make the spectrum distribution of nasal, musical and dental sounds uniform and coherent, so as to make their pronunciation smooth.

The enhancement of 1 / 2 octave narrow-band equalization is easy to make the sound source strange. Although this equalization method can greatly change the sound color of the sound source, if it is not for special effects, the equalization of singing pronunciation should be based on sound perception. In order to carry out specific effect processing on the basis of not damaging the human voice sense, 1 / 5 octave frequency range equalization processing can be used. There are the following scenes in detail.

(1) due to narrow sound sense and lack of thickness, the attenuation treatment of 1 / 5 octave can be applied at 800Hz, and the maximum attenuation value can reach - 8dB.

(2) the sound is very bright, but weak and lack of penetration. The attenuation treatment of 1 / 5 octave can be applied at 6800hz, and the maximum attenuation value is - 3dB.

(3) the rolling tooth sound has a sharp sound sense, and the "Shh" sound lacks a sense of clarity. It can be attenuated by 1 / 5 octave at 6800hz, and the maximum attenuation can reach - 6dB.

For the equalization of sound source, it is better to use the equalizer that can display the equalization curve. For example, the equalizer of all digital mixer has the function of displaying the equalization curve. In this way, the shape of the equalization curve can be seen during equalization processing, which brings convenience for future readjustment.

Generally, the equalization gain button on the equalizer of digital mixer is indicated by "g", the equalization frequency button is indicated by "F", and the equalization bandwidth button is identified by "F" or "Q".

Article source: Jiangmen recording production http://www.sjttv.com