Figure 1: Typical conference scenario using Microphone Array Beamformer
Beamforming, also known as Spatial Filtering, is a signal processing technique used in case of microphone array processing. Beamforming exploits spatial diversity of the microphones in the array to detect and extract desired source signals and suppress unwanted interference. Beamforming is used to direct and steer the composite microphones’ directivity beam in a particular direction based on the signal source direction. This technique helps to boost the composite range of microphone and increases the signal-to-noise (SNR) ratio. Figure 1 shows how Beamformed microphone array picks signal source of interest. In simple terms, the signals captured by multiple microphones are combined to produce a stronger one. To sum up and list, the advantages of Beamforming are –
- Microphone array Beamforming provides high quality and intelligible signal from desired source location while attenuating the interfering noise.
- Beamforming avoids having head-mounted or desk-stand microphones locally per speaker and also avoids any physical movement by both speaker and microphones.
- Compared to single Directional microphone, microphone array Beamforming enables us to locate and track signal source.
The above advantages render Beamforming very helpful in the field of underwater acoustics, ultrasound diagnostics, seismology, radio communications, etc. However in this particular article, we will talk about Speech enhancement using microphone array beamforming.
Figure 2: Beamformer picking signals of interest. Noise sources like projector, windows, etc. are not picked by Beamformer
In case of microphone array, there is interference between signals captured by each microphone. The microphones are placed in such a way that the resultant pattern from all the microphones allows signal energy from a particular direction. This is achieved by combining signals from one direction to undergo constructive interference and destructive interference for all other directions. This is the fundamental principle of Beamforming. Needless to say, the Beamforming combines the effect of microphone array to simulate a directional microphone whose directivity pattern can be steered dynamically based on desired signal source.
There are some products where multiple cardioid microphones are used to pick signals from different desired directions. These also provide an option to tune the directions from which we can suppress the signals, like door, window or projector, which are possible noise sources.
Following are 2 major types of Beamforming techniques -
- Fixed Beamformers
- Adaptive Beamformers
Fixed Beamformers – is the category of Beamforming techniques where signal source and noise source location is fixed with respect to microphone array. Delay-and-Sum, Filter-and-Sum, Weighted-Sum are some of the examples of Fixed Beamformers.
Adaptive Beamformers – is the category of Beamforming techniques where signal source and noise source can be moving. Making this technique useful for many applications where the Beamformer needs to adapt itself and steer in the direction of signal of interest and attenuate noise from other directions. Generalised Sidelobe Canceller (GSC), Linearly Constrained Minimum Variance (LCMV, Frost) and In situ Calibrated Microphone Array (ICMA) are some of the examples of Adaptive Beamformers.
The optimal Beamforming technique to use in a particular application largely depends on the spectral content of source signal and background noise. However in practice, this assumed information can change and in some cases the information is not available. So there is always a trade-off based on the Beamforming technique used.
Applications of BeamformingApplications of Beamforming are immense. Beamforming was initially used in military radar applications. Beamforming is now also used to capture high quality and intelligible audio signal with advances in microprocessors and digital signal processing. Beamformed signal can be used in Speech processing, Hearing aid devices, Sonar, Radar, Biomedical, Direction of signal arrival detection and aiming a camera or set of cameras towards the speakers in a video conference and Acoustic surveillance applications.
PathPartner has ported, tuned and optimized Beamforming for real-time applications like Conference speakerphones with microphone array. Challenges are immense in case of conference speakerphones in acoustic environment. Apart from microphones placement in array; wide-band speech, echo, multiple speakers, moving signal source localization and tracking are some of the major challenges to be considered. Since Beamforming focuses on a particular direction of interest, the Acoustic Echo Canceller (AEC) part of the speakerphone benefits by getting majorly signal of interest, excluding the echo paths out of the Beamforming direction as shown in figure 3.
Figure 3: Application of Beamforming along with AEC in Conference Speakerphones
Noise creeps in from all direction and can be easily removed from the signal of interest from a particular direction, thereby increasing the Signal-to-Noise ratio (SNR). Voice Activity Detector (VAD) is one of the modules which benefits from the increased SNR to a large extent.
We have designed and implemented all modules in the speech processing chain as part of the conference speakerphone. Beamformer, Acoustic Echo Canceller, Double Talk Detector, Noise Reduction, Voice Activity Detector, Non-Linear Processing, Comfort Noise Generator adapting to background noise and Automatic Gain Controller are some of the modules readily available working in real-time on Sharc DSP, Hexagon DSP, ARM and x86.
We have expertise in wide variety of other platforms for porting and optimization. Please contact email@example.com for queries.