Audio Forensics: Speech quality versus intelligibility

Engineering Technology Magazine (Institute of Engineering and Technology, UK) has a well written article concerning research by CLEAR (Center for Law Enforcement Audio Research) concerning the effect of noise reduction on intelligibility of evidentiary recordings and other audio in law enforcement situations.

... most speech-enhancement techniques improve sound quality at the expense of intelligibility, particularly when the signal-to-noise ratio (SNR) is very low. Closely related as they are, speech quality and intelligibility are not identical.

This is a classic trap that is easy for a novice to fall into.

Read it all.

(Hat tip: RH)

Filtering: Video clip explains Auto-Tune software

Nova's ScienceNOW by PBS (Public Broadcasting Service, USA) has an informative video clip on Andy Hildebrand's pitch-correction software called Auto-Tune. This audio filtering technology is used by professionals to correct mis-sung notes both in the studio and during live performances. It has been compared to Photoshop for voices.

Enjoy!

Image Restoration: iMovie Stabilization feature

Despite the best of intentions, I have not yet been able to get around to testing Apple's new iMovie video stabilization feature for the Mac OS (Operating System). I will, however, whet you appetite with a Macworld review I've come across that gives examples.

Enjoy!

Image Recognition: Military requirements push technology forward

The Strategy Page has an article about how the proliferation of video on the battlefield (e.g. from surveillance cameras and even night vision goggles) is driving technology development. Initially, it was the UK (United Kingdom) that drove image recognition technology with its massive deployment of CCTV in public spaces (e.g. train stations, airports, city centers, and shopping malls). Now, it is the military.

The benefit from computer-assistance in analyzing video comes from the following:

1. Volume (i.e. the sheer number of cameras and, hence, images to be monitored and/or analyzed)
2. Concentration (i.e. the human visual system loses the ability to concentrate effectively on images after about 20 minutes of continuous viewing)
3. Memory (i.e. computers can track and, possibly, predict more things simultaneously than a human can because, generally speaking, computers are not nearly as attention and working-memory limited in the short-term as humans and far outstrip us in their ability to recall video sequences over the long term - they can just play back the video recording off of their hard disks)
   

The article claims that the abilities of computer systems to recognize patterns is rapidly improving and approaching that of humans - that is no small feat as humans are simply amazing in their ability to perform real-time pattern analysis.

One intriguing point made in the article is that the current conflicts are generating a lot of real-world recordings of "bad behavior" that can be used to train and test new pattern analysis and prediction algorithms - training data like this is like gold to us signal processing types!

Enjoy!

Image Recognition: Military uses pattern recognition to detect bomb planting activity

I wanted to call your attention to an article by the Strategy Page on how the US military in Iraq is using pattern recognition on imagery to detect IED (Improvised Explosive Device) emplacement activities.

A very simple implementation of this technique is to take a picture on Day 1, then another picture from the same location and in the same direction on Day 2. Next subtract the two images from each other and see what is left over. All the things that did not change between the two images will disappear and only those things that changed will be left! In this case, what will be see on the "difference image" might be some disturbed gravel, tire tracks off the side of the road (assuming it is a road-side bomb emplacement) or, if the military guys are really lucky, what will be left is a bunch of guys standing around with shovels and a bomb as they work to put it in place. Neat (and helpful)!

Image Forensics: More on tamper detection

Steve Eddins (Mathworks Inc., USA), writing on his company-sponsored blog on image processing, has an excellent post regarding an article by Professor Henry Farid in the June 2008 edition of Scientific American (USA, center-left, scientific magazine). The article is primarily about tamper detection or, in other words, spotting retouching or "photoshopping" of photographs.

Image Forensics: How to spot photo retouching

Slate (US, center-left online magazine) has an article (Were the Dove ads retouched? How to spot Photoshop chicanery) on some of the techniques used in identifying if a image has been retouched. In a forensic examination of evidence, this procedure is known as tamper detection. It should be noted that the author, Jacob Leibenluft, reached out to several experts while researching the topic, including George Reis of Imaging Forensics.

Enjoy!

Image Forensics: An overview

I recently purchased and read the excellent book, Photoshop CS3 for Forensic Professionals, by George Reis and can highly recommend it. Today I came across an overview of image forensics written by him and thought I'd share it with you. The overview article, in an issue of MacTech, can be found here. Enjoy!

Miscellaneous - List of links for audio geeks

Too much work PLUS too few mentally productive hours EQUALS a list of links without any in-depth analysis. Sorry about that, but I hope to return to blogging more seriously soon. Until then, enjoy these:

Stereo from a single box (Blumlein was an audio genius. If you are an audiophile, I suggest you read a biography on him sometime.)

How the iPhone blew up the wireless industry

Acoustic superlens could mean finer ultrasound scans

Ultra-thin digital voice recorder

Noise reduction technology for cell/mobile phones (Note that the sales spiel uses words like "crystal" and "pristine" but you can clearly hear artifacting at low levels. Not trying to be overly critical, but couldn't they have let the results speak for themselves without the spin?)

Speech Recognition: Application to university lectures

Technology Review (an MIT publication) has a comprehensive article on an up-to-date application of speech recognition, including both the recognition engine and the user interface layers - namely, MIT's Lecture Browser website, a search application for video lectures.  The article is very well written and is approachable by anyone with a bit of science background.

Computer: MacBook update

In case you are interested in the experiences of a recent convert to the "Cult of Apple", my impressions of the Apple MacBook are still very positive. At this point, I have had far fewer issues with this laptop than with any previous laptop that was 'designed and built for Windows'.  I phrased it that way because, as you might recall, I am running Mac OS-X and Windows XP Pro on this machine.  I've had no significant problems with either one.

As far as tested software packages, I've extended the list of applications I've installed and tested on the Mac OS to include Matlab (a signal processing software environment useful for filtering and plotting data, including audio and video), Octave (a freeware alternative to Matlab), Remote Desktop (for connecting to a networked Windows desktop in the office), QuickSilver (a desktop search application to beat all desktop search applications), and Gimpshop (a freeware alternative to Adobe Photoshop).  I haven't fully tested all of them, but I can say that they all install and run correctly.  Octave did give a bit of trouble during installation, but once I got it to launch the first time, it seems to be behaving properly now.

Forensics: Power Line Signatures and other Audio Forensics in the news

A frequent contributor to this blog (a.a.) found this Wired article about audio forensics that came out of the annual AES  (Audio Engineering Society) convention in New York City (USA) last week.  Aural steganography (hiding data in audio files, such as MP3s), power line signature recognition, and digital editing/tampering all get mentioned.

Aside: a.a. and I have discussed the technique of power line signature recognition on several occasions.  The basic idea of this technique is that the the fundamental frequency of each local power grid varies over time. In other words, if you measured the frequency at your wall outlet, it would not read 50 Hz or 60 Hz, depending on what country you are in, exactly and constantly.  Instead, the frequency would constantly move around a little bit, perhaps as much as a few Hz up and down, all day and night, in a somewhat random fashion.  This power line "hum" leaks out too - into devices that are plugged into the power grid and even out through the air into nearby devices such as handheld audio recorders.  If one can then extract this frequency or its harmonics (i.e. multiples) from a recording and compare this "signature" to a database of the frequency variations of the same power grid, one could theoretically determine what time the recording was made.

Returning to the article, one has to wonder exactly how power line signature recognition could be used to determine the date of a recording made in a cave - do they typically wire up caves to the power grid?  I could believe that a cave would have a portable power generator, but how would the examiner get a record of its frequency variations?  Perhaps I am missing something here or the author misunderstood something from the presentation and/or interview.

Document Forensics: Recovering Archimedes' prayer book

Multi-spectral and x-ray fluorescence imaging were needed to recover the text and diagrams from a long-lost copy of Archimedes' own prayer book.  The book was in very bad condition (written on parchment that had since been re-used to make another book) and it has taken nearly a decade to recover the content. Details can be found in this Science News Online article. Fascinating stuff, not just because of the insights gained into Archimedes, but also from the laborious forensic work required to recover the content.

Audio: Build your own noise canceling headphones

Headwize has the schematics and explanation.

Computer Audio: Audition 3 in the works and an audio forensics wish list

First it was a rumor and now it is confirmed - Adobe Audition 3 is in development and Adobe plans to announce more about it later this year. A search of their web site didn't yield any additional details but there is a post on Hart Shafer's blog (Hart's Audition) that confirms it and mentions that it is on a different schedule than the rest of their Creative Suite. Hart Shafer is the product manager for Adobe's studio products, which includes Audition.

Admittedly, audio forensics is a small, niche market away from their core customer base, but if we are lucky, our needs will overlap the needs of their studio customers and a fine product will become even more useful to us. From a forensics point of view, here are the things that I have on my wish list:

• Improve support for working with multi-channel WAV files that are not intended for surround sound, including easy trimming of multiple channels and saving, for starters.
• Speed up loading of large multi-channel files
• Improve the spectrogram interface to allow easy and accurate display and analysis of specific frequencies.
• Include an option to decrease the streamlining (in other words, an "option to allow more options")
  

If you have anything to add to the list, just email me (link found on right side, near the top - don't forget to remove the anti-spam text!) or post a comment.

Video Filtering: IBM Research on Automatic Face Masking and Privacy

IBM has been moving into the video surveillance services market as of late. Now comes word on some of its research into technologies that might support that. They are not the first to work on automatic face masking but it is the first time I've seen this privacy angle to it - automatically detect, track, and mask (i.e. blur) all faces on an incoming video stream and then selectively unmask as required to identify someone. I've worked with this regularly over the last ten years, so I know that it is a very challenging problem to do reliably, particularly with "real", not laboratory, video.