TWS headset digital signal ADM
With the continuous growth of the TWS (True wireless stereo) headset market. Users’ needs for product experience have also been upgraded from simple quick links to higher standards. For example, as of this year, a large number of TWS headsets featuring clear calls have emerged on the market.
To enable clear voice communication in very noisy environments, is it possible to generate schemes that combine the signals from the inner ear and external microphones to implement an intelligent, environment-adaptive sub-band mixer technology. In fact, some domestic and foreign algorithm companies are committed to this, and have achieved certain results.
Of course, many solution companies now have special emphasis on call noise reduction solutions such as edge AI (this is one), but in fact, it is more optimized for the existing call noise reduction solutions, so this part is removed, let’s look at some basic parts first The introduction, that is, what the call noise reduction can do.
Overall, call noise reduction relies on Uplink (uplink) and Downlink (downlink) synchronization. Roughly Microphone Array/AEC/NS/EQ/AGC/DRC, the logical relationship is as follows:
ADM (Adaptive Directional Microphone Array) is a digital signal processing technology that creates a directional or noise-cancelling microphone using only two omnidirectional microphones. The ADM automatically changes its directional characteristics to provide optimal noise attenuation in a variety of environments while maintaining adequate signal quality. The adaptive process is fast, has strong frequency selectivity, and can eliminate multiple interferences simultaneously.
In addition to its good directional characteristics, ADMs are more susceptible to wind noise than traditional acoustic directional microphones. ADM technology allows two types of microphone configurations: “endfire” and “broadfire”.
In an endfire configuration, the signal source (the user’s mouth) is on the axis (the line connecting the two microphones). In a broadside configuration, it targets a straight line on the horizontal axis.
In an endfire configuration, the ADM has two modes of operation; “far-talk” and “close-talk”. In far-pass mode, the ADM acts as an optimal directional microphone, attenuating the signal from the rear and sides while preserving the signal from the front. In close-talk mode, the ADM acts as the best noise-cancelling microphone, effectively eliminating distant sounds. The relative freedom of acoustic design makes ADMs ideal for cell phones, which allow for “soft” switching between far-end speakers and near-end speakers. However, when this type of design is used on earphones, especially TWS earphones, it is more restricted by whether the user wears it correctly. Similar to airpods, the author has observed that many people have “all kinds of strange” wearing methods in the subway, some of which are the user’s ears. The shape, and some wearing habits, cause the algorithm to not necessarily work in an ideal situation.
Acoustic Echo Canceller (AEC)
When a portion of the signal in a duplex (simultaneous two-way) communication returns to the source signal, it is called “echo”. In long-distance analog and almost all digital communication systems, even small echo signals can cause interference due to severe round-trip delays.
In a voice communication terminal, acoustic echoes are generated due to the acoustic coupling between the speaker and the microphone. Due to the nonlinear processing applied in the communication channel, such as lossy vocoders and transcoding, the acoustic echoes must be locally processed (cancelled) inside the device.
Noise Suppressor (NS)
Noise suppression technology reduces stationary and transient noise in single-channel speech signals, improves signal-to-noise ratio, improves speech intelligibility, and reduces hearing fatigue.
Of course, there are many specific methods in this part, such as BF (Beamforming), or PF (Post filter) and other adjustment methods. In general, AEC, NS, BF, and PF are the core parts of call noise reduction. It is true that each algorithm solution provider has different advantages and disadvantages.
In a typical voice communication system, the level of the voice signal may vary widely due to the distance between the user and the microphone, and due to the characteristics of the communication channel.
Dynamic Range Compression (DRC) is the easiest way to equalize signal levels. Compression reduces the dynamic range of a signal by reducing (compressing) strong speech segments while sufficiently preserving weak speech segments. Therefore, the whole signal can be amplified extra so that weak signals can be heard better.
AGC technology digitally increases signal gain (amplification) when the voice signal is weak, and compresses it when the voice signal is strong. In noisy places, people tend to speak louder, and this automatically sets the microphone channel gain to a small value, thereby reducing ambient noise while keeping the voice of interest at an optimal level. Also, in a quiet environment, people speak relatively quietly so that their voices are amplified by the algorithm without too much noise.
Post time: Jun-07-2022
