The microphone array mainly simulates the auditory function of the human ear. The human ear can not only hear the sound but also have the ability to distinguish the type and direction of the sound source. In this way, we can choose to hear the sound we like. For example, in the office, when we communicate with colleagues, we ignore other surrounding sounds. In fact, the microphone array is the ear of the machine, not only collecting audio signals, but also for better sound source discrimination and noise filtering, so as to guarantee the effect of human-computer interaction.
Due to the very complex structure of the human ear, microphone arrays (especially consumer-grade microphone arrays) are difficult to achieve human ear identification and adjustment capabilities. At present, the microphone array mainly has four functions.
Sound source localization: Accurately speaking, the microphone realizes the lateral direction of the sound source rather than the precise location. Its main role is to detect the location of the sound source for subsequent beamforming. In general, sound source positioning is achieved during the wake-up phase of speech.
Noise suppression: The noise here generally refers to environmental noise, such as the sound of air-conditioning blows, the sound of cars driving outside the window, etc. Such noise usually does not cover up normal speech, but only affects the clarity of speech. Microphone arrays mainly rely on beamforming. Suppression of noise interference outside the main lobe to achieve noise suppression.
Gain adjustment: mainly to solve the problem of change in the distance of pickup, because the far field's interaction distance can be varied, so the sound source is also different in the size of the sound, such as when the person is far from the microphone or the person's voice is smaller, the microphone picks up The sound source signal will be relatively small. At this time, the signal collected by the microphone needs to be amplified to improve the accuracy of speech recognition. When a person joins a microphone to speak, or when the sound from the outside world is loud, the signal collected by the microphone is large, and even beyond the range that the microphone can collect. At this time, the microphone array is required to perform relevant processing, and the sound source is appropriately attenuated. The signal thus achieves an effective balance of picking up the sound source.
Echo cancellation: The echo here is not an echo in the traditional sense. It refers to the sound produced by the voice interactive device. For example, when the speaker plays music, the user wants to interrupt it. At this time, the echo refers to the speaker itself playing music sound. If the microphone array does not have an echo cancellation function, then the sound collected by the microphone includes the instruction sound and music sound sent by the person. In this case, it will obviously interfere with the effect of speech recognition. The purpose of echo cancellation is to eliminate music sound but retains the user's voice.
MAONO is an innovative designer and manufacturer of Lavalier, Podcasting, Wireless, Shotgun, Recording microphones and accessories for Smartphone, Camera and PC, etc.