HTC recently unveiled its latest addition to the ever-growing market of Virtual Reality headset, the Vive Pro HMD, most notably the first VR headset on the market equipped with Active Noise Cancellation (ANC). Though one tends to associate ANC with only high-end headphones, there is much scope for the technology even larger number of applications. With the noise level around us being on a constant increase, one cannot simply overlook the health hazards of continuous exposure to high levels of noise in various environments - such as boardrooms, sophisticated industrial spaces, emergency vehicles, hospitals and military vehicles.
The noise cancellation technology in general has achieved great levels of perfection when it comes to headphones and microphones used in smartphones, which do not have to deal with spatial considerations, but the real barrier appears when dealing with large area. To increase the effective zone of silence in a wide area in space has been a major challenge for researchers. Researchers are constantly looking towards developing newer methods (algorithms) and improving the existing ones to address noise in a wider area efficiently.
The biggest problem in implementing noise cancellation outside of a headset is the exponential increase in the target area. The more the number of points where noise needs to be eliminated, more is the processing power needed, and more complex the cancelling algorithm becomes.
Noise in a given area can be reduced in two ways:
1. Passive Noise Control:
In this method, the noise source is physically separated from the subject (listener) so that sound waves travelling from the source to the subject get as much attenuated as possible. The isolation can be achieved by building soundproof chambers, using sound absorbing materials like Styrofoam and jute, carpeting of floors and other methods. This method is not suitable for sounds that have a low frequency (for example the bass in music). Low frequency sounds are carried by waves of relatively lower energy and hence pass through without much attenuation. To control and reduce sounds of lower frequencies, we need active noise control.
2. Active Noise Control:
Active Noise Control (or, Active Noise Cancellation) is a newer technique that works on the principle of cancelling noise by adding (superimposing) it to its phase reversed replica. For example, if we add a sine wave to an exact replica of the same sine wave, but with a 180° shift in phase. This would cause destructive interference when superimposed and hence help in noise level reduction.
Fig. 1, Superimposition of noise and anti-noise (Wikipedia)
One of the major concerns in cancelling noise in 3-D space would be the use of a number of error microphones, cancellation speakers and adaptive filters to enhance the accuracy which would add to the system complexity. Another barrier is to account for the different phase shifts of all the frequencies as they are reflected around the room. A speaker to cancel the sound that is coming from a different location in the room could generate a different phase shift for other frequencies as compared to the source, so the cancellation would be poor. This could substantially increase the effective noise level received at some zones, instead of decreasing it because of the occurrence of constructive interference.
The possibility of this happening increases a lot when the source of the sound is at the side of the listener and not directly in front or at the back. This would generally lead to noise being cancelled at one ear, but being amplified due to constructive interference at the other ear.
Commercial applications of spatial noise cancellation including noise cancellation in aircraft cabins and car interiors are primarily based on FFT analysis owing to the cyclic nature of engine vibrations. But the real challenge is to eliminate the non-periodic ambient noise having a multitude of frequencies and phase shifts so as to create a zone of silence.
As appealing as the concept is in its application, the complexity of the science behind it is also appreciable. The most widely used algorithm for achieving active noise cancellation is the very basic Least Mean Squared (LMS) algorithm.
The LMS algorithm is one of the first and still widely used algorithms for noise cancellation. It uses an adaptive FIR filter which estimates the desired filter coefficients such that least mean of square of error signal is obtained. The filter coefficients are adjusted to minimize the error using the steepest descent method[i].
Fig. 2, Block diagram showing implementation of LMS Algorithm. (Indian Journal of Science and Technology)
The algorithm is implemented so as to reduce the amount of residual noise being received by the subject (e(n) in figure). The phase reversed anti-noise signal (y(n)) is generated in real time using a controller unit which may be a Digital Signal Processor or any other microprocessor/microcontroller programmed for the purpose. The error signal (residual noise) is often fed back to the system so as to help the algorithm adapt to the present output and change the running parameters (filter coefficients) to further decrease the residual noise level.
The LMS algorithm has been widely implemented in vehicular active noise cancellation systems. These systems employ ANC to cancel out all the noise emanating from outside the vehicle and create a zone of silence for one or more persons sitting inside the vehicle[ii].
The systems may also include a mechanism to differentiate between the following sound signals:
· Outside the defined voice band (to be eliminated) and inside the voice band (may or may not need to be eliminated)
· Emanating from outside the vehicle (to be eliminated) and from inside the vehicle (not to be eliminated)
The decisions are taken by a weighting process carried out by a discriminator which takes input from various sensors and microphones installed throughout the vehicle and decides whether the signal inputs lying in the voice band are to be cancelled or not. For example, a person’s voice coming from outside the vehicle shall be cancelled, but a person’s voice coming from inside the vehicle shall remain unaffected.
Noise reduction inside a vehicle cabin can also be achieved by reading in the road and travel conditions and using a pre-determined and stored set of filter coefficients for running the LMS algorithm[iii]. The road conditions are then constantly monitored using a variety of sensors that sense data from vehicle suspension, acceleration, engine speed, etc. Whenever a change in state is detected, the filter coefficients are dynamically changed by picking up a set of coefficients from the memory.
In another implementation, a modified LMS algorithm known as DXHS algorithm (Delayed X Harmonic Synthesis) has been used. This is specifically useful in applications where the noise source consists of only fixed harmonics, such as in ambulance sirens. This has been used to silence the noise of ambulance siren for the paramedic crew, via the use of ANC enabled headsets[iv].
A more recent enhancement to the LMS algorithm is the Filtered-X LMS (FxLMS) algorithm that further reduces the time taken to estimate and invert the noise component.Cancellation paths play a critical role in ANC systems, and the filtered-x LMS (FxLMS) algorithm takes them into account by filtering the reference signal with an estimate of the cancellation path transfer functions, which are often modelled online or at regular intervals in order to maintain the stability of the system[v].
This technology is already implemented in a wide range of ambient noise reduction headphones. Broadly there are two different arrangements used :
Feedback Arrangement (FB) : uses a noise capturing microphone inside the ear cup. It is generally implemented in headphones with large ear cups.
Feedforward Arrangement(FF): uses a noise capturing microphone outside of the earcup. It is generally implemented in ear bud style earphones.
With the advent of wireless headphones based on Bluetooth and Wifi, there is an additional battery powered circuitry for wireless reception of audio data which tends to make the system bulkier. The ANC functionality can be incorporated into the wireless communication controller so as to optimize the system.
One such implementation[vi] uses an ANC controller employing a fixed feedforward controller which obtains the error signal from an external microphone, a fixed feedback controller which obtains the error signal from an internal microphone and an adaptive feedforward controller to form a hybrid feedforward-feedback controller for attenuation of broadband noise. The coefficients of the adaptive feedforward controller are determined in accordance with the FxLMS algorithm.
Another implementation is an electronic pillow that abates snoring and other environmental noises[vii]. It uses a multichannel feedforward ANC system using adaptive FIR filters based on the FxLMS algorithm.It creates a quiet zone centred around the user by detecting noise such as snoring or other environmental noises and generating a cancelling signal using ANC. For this purpose it has multiple embedded error microphones and multiple speakers placed at predetermined positions and a controller unit coupled to them.
The advancements in technology have enabled us to see what we want to see, and even hear what we want to hear, gaining control over the otherwise involuntary sense of hearing. Algorithms that improve upon commercially used LMS algorithms (FxLMS and DXHS being just two) have been widely successful in research environments, and are waiting their turn to make a commercial impact. Yet, looking at how rapidly the consumer electronics market has grown over the last decade - and the ever increasing consumer demand for new technologies - ANC remains prime real estate for new innovation and new applications in not only consumer electronics but also in industry, transportation and healthcare.
References:
[i] Ambulance Siren Noise Reduction using LMS and FXLMS Algorithms, M. Sharma, R. Vig; Indian Journal of Science and Technology
[ii] Noise Reduction Apparatus, US Patent US007020288B1
[iii] Vibration/Noise Active Control System For Vehicles, US Patent US005758311A
[iv] An Active Control Headset For Crew Members Of Ambulance, Y .Shimada, T.Fujikawa,Y.Nishimura, T.Usagawa and M.Ebata; IEEE/TENCON 99
[v] A Review on Filtered-X LMS Algorithm, Sakshi Gaur and V. K. Gupta,International Journal of Signal Processing Systems
[vi] ANC for BT Headphones, US Patent US 2012/0170766A1
[vii] Electronic pillow for abating snoring /Environmental noises, Hands free communications,US Patent US8325934B2