tag:blogger.com,1999:blog-2569772432953120875.post4846525900374998842..comments2024-02-22T19:45:29.872-05:00Comments on Sources And Methods: Is Forensic Speaker Recognition The Next "Fingerprint?"Kristan J. Wheatonhttp://www.blogger.com/profile/02566135545863154089noreply@blogger.comBlogger4125tag:blogger.com,1999:blog-2569772432953120875.post-79861859062196415182013-08-08T10:54:54.678-04:002013-08-08T10:54:54.678-04:00http://hltcoe.jhu.edu/http://hltcoe.jhu.edu/Anonymoushttps://www.blogger.com/profile/07187789745433429425noreply@blogger.comtag:blogger.com,1999:blog-2569772432953120875.post-59898832311753130642013-08-08T08:25:53.236-04:002013-08-08T08:25:53.236-04:00CT - Fusing the results of several techniques is t...CT - Fusing the results of several techniques is the goal here. As stated, FSR is no where near accurate enough to be the only methodology behind a claim or the only piece of evidence behind a conviction. It is viable and extremely promising as a supplement to other techniques. One of its biggest drawbacks is that it isn't age-stable at all. Your voice changes drastically until after puberty, and then continues to change in minimal yet significant ways in accordance with age, health, psychological/emotional state, weather, etc. There are too many external factors that influence voice. Bottom line, it's not a fingerprint.<br /><br />RK - I agree that FSR is not perfect (hence why I try very hard in this post to present it as a supplementary methodology), but calling it a "quack" is definitely a stretch. The thing about FSR is that it encompasses so many different approaches, it is hard to boil down to one single thing. Spectrographic analysis alone is, as you say, highly inaccurate, because there are no clear cut acoustic identifiers for an individual human voice. Human expert acoustic analysis, however, yields more promising results, and automated speaker recognition technology has come a LONG way since the early 2000s (just look at work on MatLab, etc.). Ultimately, it is a methodology that, depending on how and when it is used, can contribute greatly to ongoing analysis. Equally as important as knowing when to use it is recognizing when NOT to use it (when its shortcomings outweigh its potential benefit). Melonie Richeyhttps://www.blogger.com/profile/13605783749071033745noreply@blogger.comtag:blogger.com,1999:blog-2569772432953120875.post-58083826756823853102013-08-07T15:18:19.694-04:002013-08-07T15:18:19.694-04:00Oh no Kris, please don't fall for this. It...Oh no Kris, please don't fall for this. It's well proven that voiceprints are not accurate. Spectrographic analysis only gives you a picture of what sound is being uttered. No unique acoustic identifier exists for people's voice...there's been a ton of research done on it. You can google Dr. Harry Hollien and read some of his information/research into voiceprints because he's the authority in the field. That so-called science is literally a quack. Check out this article for a start: http://news.google.com/newspapers?nid=888&dat=19750218&id=NEFSAAAAIBAJ&sjid=eHkDAAAAIBAJ&pg=7011,2840620Rachel Kesselmannoreply@blogger.comtag:blogger.com,1999:blog-2569772432953120875.post-48970975926497916252013-08-07T15:02:37.296-04:002013-08-07T15:02:37.296-04:00As a Bayesian, I'm quite happy to suggest we f...As a Bayesian, I'm quite happy to suggest we fuse the results of several techniques. :-)<br /><br />But with FSR alone, how age-stable is current voiceprint recognition compared to current fingerprint recognition?ctwardyhttps://www.blogger.com/profile/06561863222359209689noreply@blogger.com