Below are the basic categories of work I've recently been involved with recently.
The best way to describe my Web3 involvement is to provide a pointer to my most IEEE-sponsored talk on this subject:
“Of Blockchains, Crypto and NFTs: Essential Blockchain and Crypto Mechanics for Engineers”
(for video click here /
for slides click here).
This talk serves as an introduction to important Web3 technologies such as DAOs, blockchains, cryptocurrencies, NFTs and has special focus on blockchain and crypto mechanics. Consensus algorithms, random number generators, and use of cryptographic hash functions will be discussed as well as “proof-of-work” and “proof-of-stake” techniques employed by most common cryptocurrencies.
The aim of this talk is to provide a deeper level of understanding of these technologies than the overly simplistic and often inaccurate explanations of blockchain typically espoused by the media.
Developed the world's first "self-synchronizing" Discrete Sequence Spread Spectrum (DSSS) for use in highly-reverberant acoustic environments.
A design requirement was correct operation in a highly-reverberant acoustic volume in which many homogeneous ultrasound transmitters could reside. Since these transmitters use the same bandwidth at the same time - the operation required that reception of at least one of the transmitters be decoded. A form of DSSS was designed - as SS allows for reception in negative SNR (interference from other transmitters is "noise").
This system uses the same set of codes for both (DSSS) pilot and user information - allowing both to be perfectly orthogonal at the transmitter. The codes chosen have a N*log2(N) decomposition, allowing fast computation and (even more importantly) fast cross-correlation at the receiver (needed for synchronization).
This system - when decoding at its limit in negative SNR - is within a factor of 10x of the Shannon bound. This is even more impressive since room reverberation represents a very severe form of multi-path. Six patents have been filed from this work and this technology has been incorporated into product; the base patent is US 10,003,377, some of the remaining 5 can be found at the patent and papers link at this site.
Acoustic ranging is also possible with the resulting design. A highly flexible and extensible transmit format has been designed and allows for sending information elements such as IP addresses/ports, BSSIDs/passowrds and information commonly transmitted via bluetooth or NFC systems.
Many times people attempt to join web-based conferences in which the audio is already present in the room. If one is not careful to join muted, the dreaded howling/echo ensues. One way to preemptively avoid this is to sense this condition and take proactive action to avoid it. Although conceptually straightforward, unknown distortions (e.g., poor voice encoding/decoding) and room reverberations make this task more complex than it seems. Prototype in MATLAB has been developed with 100% classification success.
Face recognition (CNNs), speaker identification (RNNs), noise reduction (RNNs), cochlear modelling for speaker identification (RNNs) and duplicated Baidu speech research results (Udacity class). This work is in-progress.
The field of noise reduction via traditional techniques is well documented. However machine learning is being applied to this field which yields improved results relative to the traditional techniques. This work is in-progress.
I - along with my Cisco colleagues - developed rate adaptation algorithms and technology suitable for real-time communications (e.g., voice and video over RTCWeb ) and ultra-low delay applications (such as financial trading transactions) which must run over the best-effort Internet. I have created a large-scale laboratory networking environment in which to assess rate adaption candidate performance and to further improve performance in our future products. I have built and designed a " RMCAT LAB " by specifying software framework, media palyoads, and real-world hardware routing topology for the purposes of evaluatng video rate control algorithms. This standards work for this activity can be found at the patent and papers link at this site.
In a related technology area I have also invented and designed Delay-Minimization Technologies (DMT) that are essential for the enablement of cloud-based, real-time services which are delivered "over-the-cloud" but within larger traffic aggregates that have loss-based congestion control at the transport layer. These DMT technologies help transform ordinary Internet Service Provider (ISP) access link performance toward the real-time performance of more expensive service provider arrangements (e.g., MPLS). In doing so, they help to drive down access costs for cloud-based real-time services. See US 9,948,561 for the base patent for this technology.
I co-authored IETF transport protocol RFC 3758 (Stream Control Transmission Protocol - Partial Reliability Extension). SCTP-PR is an extension to IETF Transport Protocol RFC 2960Stream Control Transmission Protocol (SCTP) which adds "partial reliability" to a given stream in SCTP. The congetion control of the vast majority of internet packets is throttled by packet losses in what is called the "bottleneck link" - so packet losses (on the order of 1%) are expected and should be accommodated in any delay-sensitive data transmission for fog computing and "big data". SCTP-PR essentially recognizes that recovery from lost packets (e.g., packet drops) takes time. During the re-transmit interval the dropped data may become "stale" in the sense that more current information may beat the stale information to the destination because the more current information was not dropped by the network. In this scenario not only is stale information delivered late - but it was retransmitted to do so! This retransmission takes up additional bandwidth further contributing to the congestion on the link. SCTP-PR provides the application the ability to say, in effect, don't try anymore to retransmit this information if it has not already arrived. This is a highly desired capability in the world of big data and there has been recent use of this mechanism in IoT and Fog Computing environments and papers.
I brought prototypes of the lossless and stateless compression of G.711 to the ITU-T in 2007 (see G.711.0 link on toolbar). They really liked the concept and put it on a path for standardization. I moderated the resulting work item. The result was the G.711.0 codec - the first lossless codec intended for use with RTP. It is the first and only G-series codec contribution from Cisco. I am in the process of finishing the IETF RTP payload format draft for it in the IETF PAYLOAD working group as well as supporting internal business units that are implementing it.
Sought-after expert on two patent committees (voice and video and collaboration) for patents related to media quality, FEC, signal processing and communications technology.
Michael A. Ramalho
Last Modified on :
June 28, 2022
Page Owner :
Michael A. Ramalho, Ph.D.