A method and a device for suppressing noise of a loudspeaker

Abstract

The embodiment of the invention provides a method and a device for suppressing noise of a loudspeaker. The contribution degree of each frequency characteristic to the compression strength calculationresult is adjusted according to the relation between the strength of the audio signal on the frequency band capable of exciting noise and the compression strength and the relation between the strengthof the audio signal on the frequency band capable of covering noise and the negative compression strength, and original audio information is reserved to the maximum extent while noise is restrained;The method has the advantages of simple operation, low sound quality loss and low software and hardware expenditure, and can retain the information of the original audio signal to the maximum extent on the basis of effectively improving the problem of excitation noise caused by stronger signals in a certain frequency band.

Description

technical field [0001] Embodiments of the present invention relate to the technical field of loudspeakers, and more particularly, to a method and device for suppressing noise from loudspeakers. Background technique [0002] With the development of mobile and portable smart devices, people have higher and higher requirements for the audio playback quality of portable smart devices. However, due to the size and structure of portable devices and the defects of the speakers themselves, noises often appear in the playback audio. There are many reasons for noise, the most common of which is that when the energy of certain frequency bands in the sound source is high, even if the amplitude of the speaker diaphragm does not exceed the maximum design amplitude of the speaker, obvious noise can still be heard. [0003] In the prior art, there are mainly three methods to solve the above-mentioned noise problem. The first method is to compress the energy of the frequency band that will g...

AI Technical Summary

Problems solved by technology

[0004] The defect of the above-mentioned first method is that when there is a possibility of stimulating noise in a certain frequency band, the energy of this frequency band will always be attenuated, no matter whether the energy of this frequency band is really enough to stimulate the human ear. The level of the perceived noise will cause a loss in the playback sound quality. For example, when the frequency range is 600-2000hz, this approach will directly affect the playback effect of the human voice; the above-mentioned second method can stimulate the noise by detecting The amount of energy on the frequency band determines the strength of the compression applied to the frequency band. It can be done without compressing the frequency band when the energy is small; but this method will still cause a large loss of sound quality. The reasons are as follows: According to the experiment, When the original sound source has a certain amount of energy in the frequency band where the excited noise is located, the noise is easily integrated with the energy of the audio itself. At this time, it is difficult for the human ear to hear the noise; so in this case In this case, even if there is a large amount of energy in a specific frequency band, it is still possible not to compress the energy to obtain better playback sound quality; in the third method above, the defect of the first judgment mechanism is: when the high frequency in the sound source Even if the low-frequency components are very small, the ratio of high-frequency components to low-frequency components may still trigger the threshold set by it, and then activate the noise cancellation mechanism; however, the overall energy of the audio is low at this time, and cannot produce audible Therefore, it is meaningless to activate the noise cancellation mechanism at this time, and it will only damage the playback sound quality; and the noise caused by data containing piano noise has a certain frequency range. If the high-frequency components of the audio data itself If it is larger, the noise cancellation mechanism will not be triggered, but if the frequency of the high-frequency component does not coincide with or be close to the frequency of the excited noise, the energy of the original audio data will not be able to cover the energy of the noise well, thus Noise can still be heard; the defect of Judgment Mechanism 2 is: if less frequency features are considered, for example, only the ratio of high-frequency components to low-frequency components is considered, an accurate classifier cannot be trained, as described in Judgment Mechanism 1 Defects: To train an accurate classifier, a large number of features need to be considered, resulting in a significant increase in the complexity of engineering implementation and the overhead occupied by actual operation, which is not practical on mobile portable devices that are extremely sensitive to power consumption; And the judgment mechanism 2 when training SVM, if you want to get an accurate classifier, there will be problems such as long training time, large training samples, differences in training results, and uncontrollable training results.

Images