| Performance Characteristics | Silencer Devices | LMS Filters (“ANC”) | AI Spectral Subtraction |
|---|---|---|---|
| Maximum Frequency which can be handled (people can hear up to 20kHz) | GHz range (kHz x 10^6) | < = 2.5 kHz | < = 16 kHz (at Fs=32kHz) |
| Latency (delay due to processing, etc.) | Shortest | Too long for high frequencies | Excessive |
| Local Processing Power Required | Least | Reasonable | Very Significant |
| Local Memory Required | Reasonable | Least | Very Large |
| Network / Cloud Access Required | No | No | Yes |
| Suitability for Applications | Silencer Devices | LMS Filters (“ANC”) | AI Spectral Subtraction |
|---|---|---|---|
| Can be used in low cost systems | Yes | Yes | No |
| Can run at the speed of sound (for real-time suppression) | Yes | Yes, with limited frequency range | No |
| Must be positioned between noise and cancellation point | Yes | Yes | No |
| Can be effective and used in Mobile Phones for Transmit & Receive | Yes | No | Yes, if a long response delay is acceptable |
| Can be effective and used in Hearables | Yes | Yes, with limited frequency range | No |
| Can facilitate a complete redesign of Hearables (no need to block the ears) | Yes | No | No |
| Can be used to clean noise from speech for AI-LLMs, etc. | Yes | No | Yes, if the cost and latency are acceptable |
| Can be used for reducing noise in conference calls, etc. | Yes, clears my voice for others | No | Yes, clears noise from all parties for me |
| Can be used so a Voice Assistant can hear you over the music it is playing | Yes | No | Yes, if a long response delay is acceptable |
| Can be used to eliminate the Voice Assistant “self-wake” problem | Yes | No | No |
| Can be used to eliminate “Broadcast Echo” in Gaming & Virtual Environments when a user is not using a headset | Yes | No | No |
| Can be used for vibration control and low frequency engine noise | Yes | Yes | No |
| Potential for use in Car, Truck, and Plane Cabins | Yes | Yes, with limited frequency range | No |
| Potential for a Window-Mounted unit to quiet street noise in bedrooms | Yes | No | No |
| Potential for use in residential and commercial appliances & equipment | Yes | No | No |
| Potential for use in Industrial equipment and construction sites | Yes | Yes, with limited frequency range | No |
| Potential for use in terrestrial and non-terrestrial communications | Yes | No | No |
| Potential for use in Military stealth and comms | Yes | Yes, with limited frequency range | No |
LMS Filters (aka ANC):
This noise cancelling technology was invented over 40 years ago, and even with the relatively recent enhancements via predictive filters, can only cover a small portion (less than 15%) of the range of human hearing in short-distance applications such as mobile phones, hearables (earbuds, headphones, headsets, etc.), voice assistants, etc. This is because LMS systems attempt to match an audio waveform by solving a set of complex non-linear equations, repeatedly within discrete chunks of time defined by the number of samples in a chunk / the sampling rate. Therefore, passive means must be utilized to block the other 85% of the noise from reaching your ears in those common use cases, and electronic techniques like beam-forming or special microphones are required to optimize the signal to noise ratio in a microphone. It is probable that the introduction of Silencer Devices technology may make LMS obsolete for most audio applications. However, LMS-based systems may retain a place in low-cost audio products where the ear can also be blocked, and low frequency applications, such as vibration control systems.
AI Spectral Subtraction:
The goal of the AI noise reduction systems we are aware of is to extract clean speech from noisy situations, such as a VoIP or video conference call when many of the participants are working remotely. A significant advantage of this approach over the other technologies discussed in this paper is it does not need to be positioned between the noise source and the listener because it is not designed to provide anti-noise faster than the speed of sound in the real world (it is effectively acting on recordings of voice in the form of 10ms frames, and transmitting cleaner recordings of the voice). On the other hand, this technology has a very long latency (takes a long time to do the processing), requires significant local device memory and processing power, and needs consistent and robust cloud network access. These attributes make it poorly suited for use in many mobile devices, hearables, voice assistants, kiosks and other applications which require fast results and/or have processing power, memory, battery power, internet access or cost constraints.
Silencer Devices’ Technology (SDT):
This noise canceling technology is capable of extremely low latency (on the order of 1 sample for algorithmic processing in run mode, excluding any data conversion time) and extremely high frequencies, way beyond the 20 kHz audio spectrum, into the MHz, and likely even the GHz range. The technology was designed so it can be scaled to Performance Requirements and/or Cost Constraints. Run Mode requires relatively few onboard resources (simply need the ability to calculate dot products with a set of pre-loaded coefficients) because the complex calculations are done in a Calibration mode, and that process can be done completely off-device. At the current state of development, exceptional results are expected in many applications such as mobile phones, hearables, voice assistants, Kiosks using speech to text to access AI Large Language Models, PC voice control and data entry, Gaming and Virtual Environments, to name a few. Many more applications will become practical when the AI-enhanced feedback algorithm is implemented. These include applications in vehicle cabins and rooms; along with additional commercial, industrial and military applications involving stealth, large spaces and/or high amplitudes. In addition, the Feedback System will enable automatic calibration for devices using our technology, facilitate user options for performance customization, and support integration with multi-point sound systems for wide-area applications (construction sites, etc.).