MUTE: Bringing IoT to Noise Cancellation [SIGCOMM 2018]
Is this going to be a Bose killer?
Our goal is to provide an easy, general and high-level context of this paper’s contributions and our take on the implications of the paper. Please refer to the actual paper for more details if this post interests you and please feel free to contact us for errors, changes and suggestions!
MUTE [paper] is one the coolest paper in SIGCOMM 2018. If you are a frequent traveller, I am sure you have, or at least thought about, purchasing a noise cancelling headphone, such as Bose QC35 or Sony WH1000XM2. These headphones are called an Active Noise Control (ANC) devices, that works solely through acoustic manipulations, which I will go over in the next section. MUTE puts an ingenious twist on how ANC headphones work by augmenting ANC algorithm with network signals. Let’s first cover some background knowledge required for this paper and then talk about the contributions of this paper.
Background
Let’s first go over what Active Noise Control (ANC) is in detail and the current state of ANC devices. ANC is a method of cancelling unwanted sound by producing a cancelling noise, or what is called an anti-noise (Fig. 1).
Here is a very brief overview of why this works. Sound travels as a waveform, or more specifically, a pressure wave that creates regions of high pressured particles (compression) and regions of low pressured particles (rarefaction). It needs a medium to go through, such as air or water, and is very similar to an earthquake wave that travels through ground. Now, if you create a pressure wave that has an opposite phase, or in other words, a wave with an opposite up and down pattern, and combine it with the original wave, the resulting wave will be quite close to a standing wave. So, ANC devices can choose which sound (or pressure wave) to cancel out, by producing anti-noise.
In order to produce an anti-noise, ANC devices need a microphone, called a reference microphone, to absorb surrounding sound and an additional anti-noise speaker to produce the anti-noise sound. Once the microphone absorbs the sound, there is usually a Digital Signal Processing (DSP) chip that processes the sound and generates the anti-noise sound. If you have tried out an ANC headphone, a low humming noise that you hear when you turn it on is the anti-noise generated by the anti-noise speaker.
Now, a characteristic of an ANC device is that , it works really well for predictable low frequency noises, such as engine hum sounds, but does not really work for high frequency or unpredictable noise, such as human conversations. The reason why it doesn’t work well for high frequency noise is due to how close the reference microphone, DSP chip and the anti-noise speaker are to your ears. ANC devices must produce anti-noise faster than the time it takes for the actual noise to reach your ears (which is around 30 microseconds), or you will hear the original noise. However, today’s ANC devices take around 3 times more time to produce anti-noise. So for high frequency noise with no patterns, the noise cannot be cancelled in time, where as low frequency repeating sounds can be cancelled, as the noise that the reference microphone has received is similar to the noise that is about to arrive in the future. One way to mitigate this problem is to create headphones that cover your ears, so that you hear less of the surrounding noise in all conditions. Although this makes it quite uncomfortable to wear headphones, I think most people prefer this to painful airplane noises.
Which Problems is this Paper Trying to Solve?
This paper wants improve ANC further. As mentioned above, ANC headphones with reference microphones does not work well for unpredictable and or high frequency noises. As an added bonus, MUTE wants to make ANC headphones cheaper. Since ANC headphones require a reference microphone, DSP chip and an anti-noise speaker, they are quite expensive to make. For example, a Bose QC35 can cost around $350 a pop (I am sure this figure includes some extra costs for Bose branding, marketing, patent rights and things like that…). Finally, MUTE solves what was once thought as an acoustic-only problem with the newly available tool called WiFi.
How is this Paper Solving these Problems?
MUTE proposes installing an external microphones that sit closer to the noise sources. Once the noise is obtained by the microphone, it sends the noise through the WiFi signal, which travels faster than sound, and allows the DSP board to compute the anti-noise. This allows the DSP board to obtain a lookahead for future sound. Aside from the system itself, MUTE also proposes a novel algorithm called Lookahead Aware ANC algorithm (LANC) that takes advantages of the added timing benefits. The details of LANC can be found in the paper and it may need a quite a bit of background knowledge to completely understand that. As a brief overview, LANC works better if the microphone is closer to the sound source and the headphone is far away from the sound source.
What are the Results?
The results are quite outstanding. MUTE can cancel sounds at higher frequencies better than existing ANC headphone (Bose QC35), and on average, the sound cancellation is about 8.9dB better with correct noise suppression. They also tested on human subjects, who consistently told them MUTE was better in sound cancellation.
My Take on this Paper
This is a quite interesting paper that has a real practical use. However, I think this is paper is limited to use cases where the listener is sits in an enclosed space and somewhat stationary. I think this will work really well inside an airplane or a train, or when someone is sitting in a room. Also, we all know that constant use of WiFi consumes battery, yet this paper did not compare overall power usage between two approaches. Many people want ANC headphones that last a long time, so it would be great to see the comparison. Overall, I really enjoyed reading this paper and I hope this project actually commercializes!
We are looking for passionate writers from all fields of CS!
I believe that one of the main and most important habit that a grad student must have is to read papers at a regular fashion. I thought that being a paper reviewer with a light amount of responsibility gives incentives to read more papers. So, I personally started this blog for me to record what I was reading and I found it really really helpful. I hope that more can join my experience and have a great learning experience in the process! Feel free to email me at yo2seol@cs.stanford.edu if interest!
