Can anyone tell me what were the improvements in the MELPe's synthesis vs. MELP's? (they are different, i.e. different parameter interpolation and synthesis. But what is the key advantage/improvement?) How does that improve quality?
The improvements of MELPe's synthesis vs. MELP's
Started by ●February 16, 2006
Reply by ●February 16, 20062006-02-16
Kirkland wrote:> Can anyone tell me what were the improvements in the MELPe's synthesis > vs. MELP's? > (they are different, i.e. different parameter interpolation and > synthesis. But what is the key advantage/improvement?) > How does that improve quality?Good luck! A couple of years ago I tried, through a series of emails, to get better technical descriptions and a list of patents that applied to MELP and MELPe, but to no avail! I mainly wanted to know WHAT I could or could not use from MELP without violating someone's IP, but I guess that would be too easy.... -- Phil Frisbie, Jr. Hawk Software http://www.hawksoft.com
Reply by ●February 16, 20062006-02-16
Phil Frisbie, Jr. wrote:> Kirkland wrote: > >> Can anyone tell me what were the improvements in the MELPe's synthesis >> vs. MELP's? >> (they are different, i.e. different parameter interpolation and >> synthesis. But what is the key advantage/improvement?) >> How does that improve quality? > > Good luck! > > A couple of years ago I tried, through a series of emails, to get better > technical descriptions and a list of patents that applied to MELP and > MELPe, but to no avail! I mainly wanted to know WHAT I could or could > not use from MELP without violating someone's IP, but I guess that would > be too easy....I was able to get that information from Compandent (highly knowledgeable of all MELP & MELPe issues), you may try them as well at: http://www.MELPe.com http://www.Compandent.com MELP: 1. TI's patents in MELP (are carried also to MELPe). Some patents are obvious, and some are more tricky, for example there may be a relation between MELP's dispersion phase filter and post-filter's patents. TI may waive payment for its IPR as long as the implementation is on TI's DSP. MELPe: 2. Microsoft claims to have several pending U.S. and international patents in MELPe: http://www.microsoft.com/mscorp/ip/tech/melpe.asp You may contact them and ask them for the specific patent/application numbers. 3. Compandent claims to have IPR in the MELPe standard to which they have contributed as well. 4. ATT/Lucent claim to have patents in the MELPe's Noise Pre-processor: http://www.att.com/attlabs/products/portfolio.html http://www.att.com/attlabs/docs/att_melpe_factsheet_071405.pdf 5. Thales may have some IPR in the 600 bps MELPe. If I were you I might hire someone to conduct professional IP/prior-art search for me. :)
Reply by ●February 16, 20062006-02-16
>If I were you I might hire someone to conduct professional IP/prior-art >search for me. :)Here is somewhat unprofessional but free search result :) The only US patent which Compandent can possibly claim is 6,614,370 "Redundant compression techniques..." but I have no idea how it applies to MELPe standard. MS patent(s) must be coming from somewhere else, probably acquired during a process of some other company's acqusition. TI's basic patents on MELP are well known. You can browse them all at your pleasure at www.uspto.gov/patft
Reply by ●February 16, 20062006-02-16
vocoder wrote:> Phil Frisbie, Jr. wrote: > >> Kirkland wrote: > I was able to get that information from Compandent (highly knowledgeable > of all MELP & MELPe issues), you may try them as well at: > http://www.MELPe.com > http://www.Compandent.comWell, perhaps they saw my request as "I need to know what I can use in a free and open source code codec", so they were less than helpful ;)> MELP: > 1. TI's patents in MELP (are carried also to MELPe). > Some patents are obvious, and some are more tricky, for example there > may be a relation between MELP's dispersion phase filter and > post-filter's patents. TI may waive payment for its IPR as long as the > implementation is on TI's DSP.Yes, this is EXACTLY what I mean. The interpretation of a patent and the implementation can be hard to connect. I decided to avoid any MELP specific code and algorithms to be safe, but it should not be this hard to identify IP in code.> If I were you I might hire someone to conduct professional IP/prior-art > search for me. :)Yes, but for small developers that can actually HURT later on. If I hire a professional to look for IP conflicts in my own algorithms, find none, and am later sued, it could be argued I SHOULD have found the conflict and therefor must have ignored it and am guilty of willful infringement. This has now strayed off topic, and I will discuss it no further. -- Phil Frisbie, Jr. Hawk Software http://www.hawksoft.com
Reply by ●February 16, 20062006-02-16
fizteh89 wrote:>> If I were you I might hire someone to conduct professional IP/prior-art >> search for me. :) > > Here is somewhat unprofessional but free search result :) > > The only US patent which Compandent can possibly claim is 6,614,370 > "Redundant compression techniques..." > but I have no idea how it applies to MELPe standard.I think you may be assuming that IPR and patents are the same things... indeed "unprofessional but free search" So IP includes patents but also copyrights and trade secrets. For example, since Compandent contributed to the MELPe's code and someone copies its source code or a derivative and commercializes that, then he infringes on Compandent's IP... Since they have announced that on their web site (see: Note on http://www.MELPe.com ), then doing so becomes willful infringement which automatically triples the damages... Also, since you have mentioned one patent by Compandent, are you aware of all their patents or pending patents...???? indeed "unprofessional but free search"> MS patent(s) must be coming from somewhere else, probably acquired > during a process of some other company's acqusition.Microsoft indeed acquired SignalCom and its IPR, and SignalCom passed MELPe to Compandent to complete the job and continue subsequent MELPe related projects.> TI's basic patents on MELP are well known.Right, but someone still asked about them.> You can browse them all at your pleasure at www.uspto.gov/patftNo need thank you, I've already done that years ago... :)
Reply by ●February 16, 20062006-02-16
Phil Frisbie, Jr. wrote:> vocoder wrote: > >> Phil Frisbie, Jr. wrote: >> >>> Kirkland wrote: >> I was able to get that information from Compandent (highly >> knowledgeable of all MELP & MELPe issues), you may try them as well at: >> http://www.MELPe.com >> http://www.Compandent.com > > Well, perhaps they saw my request as "I need to know what I can use in a > free and open source code codec", so they were less than helpful ;)I guess that how it is when you run a business...>> MELP: >> 1. TI's patents in MELP (are carried also to MELPe). >> Some patents are obvious, and some are more tricky, for example there >> may be a relation between MELP's dispersion phase filter and >> post-filter's patents. TI may waive payment for its IPR as long as the >> implementation is on TI's DSP. > > Yes, this is EXACTLY what I mean. The interpretation of a patent and the > implementation can be hard to connect. I decided to avoid any MELP > specific code and algorithms to be safe, but it should not be this hard > to identify IP in code.Well, almost any standard vocoder or published paper is related to some patents (or other IPR), so you should avoid almost anything... :)>> If I were you I might hire someone to conduct professional >> IP/prior-art search for me. :) > > Yes, but for small developers that can actually HURT later on. If I hire > a professional to look for IP conflicts in my own algorithms, find none, > and am later sued, it could be argued I SHOULD have found the conflict > and therefor must have ignored it and am guilty of willful infringement.The exact opposite! If you hired some expert who determined that you do not infringe, then that may help you later to claim that it was not willful because of the expert opinion and you took reasonable precaution! That's why many companies spend $$$$ on such professional prior art search and expert opinion! If you do not do that, and just flatly ignore other's IPR then it will be much easier to claim that you did that willfully! BTW, TI mentions its IPR (i.e. copyrights etc.) in the MELP source code. MS, ATT Compandent & Thales mention theirs IPR in the MELPe documentations and source code as well as web sites, so if you ever think about not checking the patents or commercialize the source code or its derivative and later on claim that it was not willful infringement then your chances to get away with that approach zero! If for example you "only" sell DSP boards and accompanied them with "free" MELPe (or for low price) then you still commercialize MELPe and willfully infringe! (at least contributory infringement) Remember: Willful infringement = triple the damages!> This has now strayed off topic, and I will discuss it no further.OK, twing!
Reply by ●February 16, 20062006-02-16
Kirkland wrote:> Can anyone tell me what were the improvements in the MELPe's synthesis > vs. MELP's? > (they are different, i.e. different parameter interpolation and > synthesis. But what is the key advantage/improvement?) > How does that improve quality?MELPe uses a new harmonic synthesis method that generates the mixed signal in the frequency domain, rather than the sum of two time-domain filtered excitations as in MELP. I will leave it to others to elaborate on that and the advantages.
Reply by ●February 17, 20062006-02-17
You seem to be too obsessed with the word "professional"... But sometimes one amateur puts hundreds of "professionals" out of work... "Overspecialization breeds in weakness, it's a slow death" :-)
Reply by ●February 17, 20062006-02-17
fizteh89 wrote:> You seem to be too obsessed with the word "professional"... > > But sometimes one amateur puts hundreds of "professionals" out of > work... > > "Overspecialization breeds in weakness, it's a slow death" :-)and you seems to be too obsessed with the word "unprofessional"... So, what does that qualify you for? :) BTW, I was also obsessed with a little blond from Cherry Hill named Elaine, does that count? Not mentioning Karen Alpert from Cherry Hill who also did great job... and she was obsessed too... Now I'm too obsessed with the word "obsessed"... and now I'm not! :)
