SM
SM
This legal, non-criminal website has opinion, analysis, and publicly available information about government officers and others. This page discusses biometric identification.
C.O.s spend much time counting prisoners. When there's a serious crime in a jail, investigators sometimes wish they knew everyone in that part of the jail in the hours preceding the crime. The government sometimes wants to know where former prisoners (for example, people on parole) are. The government sometimes tries to find former prisoners; for example, people who have escaped from prison. Below we describe a system for automtically identifying prisoners as they pass various points in a jail, and for identifying former prisoners as they pass various points in public places.
The system can also be used to detect intruders, as briefly mentioned toward the bottom of this page.
Prisoners are often counted; for example, as they enter a cafeteria. It would be convenient for prisoners entering a cafeteria to get food as anyone else might, without being required to say their name to a C.O., touch a fingerprint scanner, or submit to any other identification procedure. This could be done using multiple biometry.
First, the staff would enter information about each prisoner into the system. The information would include height; weight; front and profile photographs taken at a wide variety of angles and distances using visible spectrum light and perhaps also infrared light, and a voice sample. Later, when a prisoner walks into a prison cafeteria:
Using essentially the same technique, one could measure the length of the stride (step length). A computer system observing people walk near a well could be programmed to calculate stride length. Some people have a longer stride (step) than others. One can have markers on a wall near a floor.
Another way is to have a camera observe where people walk, then have a computer analyze the data to calculate the length of the stride.
As a prisoner walks through a hallway, a light shines from one side of the hallway toward the other. That other side has many light receptors. The walking prisoner briefly blocks light to some receptors. The receptors are connected to a computer which is able to infer the prisoner's profile. The computer could recognize some anatomic features; for example: a big, protruding belly; a skinny prisoner; a head shaved bald; a head with scalp hair in a pony tail; a thin neck. In some jails, prisoners wear varied clothing. For example, some prisoners wear sandals, others wear sneakers, others wear boots. Some prisoners wear tight clothing. Other prisoners wear loose, even baggy clothing. Profiling might recognize some kinds of clothing, which could help to recognize some prisoners.
As a person walks, his hands swing back and forth. Standing prisoners vary in how close their hands are to the ground. Even when one controls for height and sex, arm length varies. As two prisoners walk, their hands may move at different heights from the ground. Profiling might help detect how far the hands are from the ground. This helps to identify prisoners.
In some jails, prisoners are allowed to smoke. Smoke, even from cigarettes, can be detected by smoke detectors. At some prisoner-identification places in a jail, smoke detectors could detect smoke from cigarttes and other tobacco. For example, as a prisoner walks past a place in a hallway, a smoke detector might briefly (for a few seconds or several seconds as the prisoner passes) detect smoke and report that to a computer. Some prisoners smoke, others don't. Those smoke detectors would not be used by the jail to detect fire. They would not set off a fire alarm.
People vary in body temperature. Infrared may be useful even if all it does is distinguish exceptionally warm people (much infrared emitted) from exceptionally cool people. It is not necessary that each person be identified by a unique, infrared image.
Standing prisoners vary in how close their hands are to the ground. Even when one controls for height and sex, arm length varies. Infrared might be able to detect how far the hands are from the ground, because hands often are ungloved even when other parts fo t he body are clothed. This helps to identify prisoners.
The tops of prisoners' heads could be observed by infrared. We guess that a bald prisoner emits more infrared energy from the top of his head than does a prisoner with a full scalp of hair. A biometric system could sometimes spot bald prisoners.
In some jails, prisoners vary in how they dress. Some prisoners usually wear short sleeves and other prisoenrs usually wear long sleeves, for example. Some prisoners wear an undershirt alone over the chest. Other prisoners usually wear an undershirt and a shirt over that. Infrared might be able to distinguish those prisoners.
All of the biometric identification about each entering prisoner goes into a computer which then immediately concludes who the prisoner is. Some prisoners may sometimes be identified using just one or two criteria (for example, facial structure identification). However, that's not necessary. It is sufficient if the system, using all information available, identifies a prisoner. Maybe his height narrows the number of possible prisoners in that prison to 100 prisoners. Maybe his weight further narrows it down to 60. Maybe facial structure identification, using visible spectrum light (not infrared), further narrows it down to only one prisoner. In such a case, there would be no need to consider other criteria such as his eye color or the pattern of infrared energy he emits.
If the computer still doesn't know who the prisoner is, the computer system asks the prisoner his name. Biometry system microphones listen to his answer, then use his voice (in addition to the other identifiers such as height) to identify him. If the computer still doesn't know who he is, the computer tells a C.O. nearby. The C.O. finds out, then enters the information into the computer system so that the computer will, in the futurre, be able to recognize that prisoner. The system can learn from experience. For example, the computer's ability to recognize a face improves much when the computer uses an average of photographs of that face, taken under a variety of lighting conditions. We discuss this and provide a scholarly citation in our Facial Recognition paragraph, above.
It would not be necessary to use, or even try using, all of the biometric traits mentioned above.
Consider a prisoner some of whose identifiers gradually change. For example, he gradually gains weight. The computer system should be programmed to remember gradual change in a prisoner's identifiers. Each time a prisoner enters a cafeteria, his weight should be compared to his most recently observed weights (for example, his weight the previous time he entered the cafeteria), not necessarily his weight on the day he became a prisoner.
We guess that height varies. A person may stand a little taller some times than other times. The biometric system would measure real height as it actually is during normal activity such as entering a cafeteria. Maybe some people walk slightly taller in the morning or after a cup of coffee, for example. If a prisoner enters prison at age 18, he may be taller when he leaves prison a few years later. The biometric system would record actual heights and use them to recognize.
A prisoner's face may change while in prison. He may grow a mustache or beard, for example. The computer system will save many photographs of him and use an average of them. This will help the system automatically recognize him after he leaves prison.
The unobtrusive, multiple-identifier system described above could be in several places in a prison: a cafteria, a hallway or gateway through which prisoners must pass to leave the prison, an infirmary, a work area, and a recreational area. A biometric system could track all prisoners as they moved through a prison every day.
The system described above would be expensive to use in prison alone. The system might be used in non-prison settings to identify former prisoners who are sought (a prisoner who escapes from prison, someone on parole who is wanted but cannot be found, or some other ex-prisoner who is sought). In the remainder of this discussion, a "subject" is a person who is sought. The biometric system could be installed in hallways and gateways in airports, office buildings, athletic stadiums,and other places. Such a system could store information about a million or more subjects.
The system could also be used to unobtrusively monitor movements of former prisoners or people on probation even if they are not sought. For example, consider an ex-prisoner on parole who goes to an airport or railroad station where he takes an escalator from one level to another. The biometric system automatically identifies him, photographs him, and sends a fax, with the photograph, to his parole officer just in case his parole officer is interested.
A false positive is mistakenly thinking the subject is there although he isn't. The system can be programmed to modify the rate of false positives for each subject. For example, consider a parole officer who is mildly curious about the movements of one parolee (subject 1) but who does not want to be bothered by many faxes about people who resemble subject 1. The computer system's file for subject 1 could be configured to require a high level of confidence (in other words, a very close resemblence) for that subject before a fax is sent (resulting in few, false positives about him). The system would conclude that it had spottted subject 1 only if it were highly likely that it had spotted him. However, the same parole officer urgently seaks a different parolee, subject 2. The biometric system's file for that parolee could be set to a lower confidence level, because the parole officer may be willing to read many false positive faxes (faxes about people who resemble subject 2 but are not subject 2) in the hope of getting a fax that really is about subject 2.
An advantage of the multiple-identifier biometric system described above is the wealth of data it stores about each subject. Consider a subject who was a prisoner for five years. The system recorded much information about him (for example: weight and skin darkness) every time he passed one of its checkpoints (for example, a checkpoint in a prison cafeteria). In the summer, his skin may have been tanned but, in the winter, his skin may have been pale. There may have been periods in which he shaved daily, other periods in which he shaved once or twice weekly, and periods in which he did not shave. He may have entered prison slim and left it fat. The computer system has five years of detailed information about him in all of those variations. When he walks through an airport corridor or goes through some other checkpoint out of prison, he could be observed and compared to all of the years of data, not just data for the day he entered or left prison. The biometric computer system has known him for years and remembers many days (maybe every day). In prison, there is little or no need to compare a prisoner entering a cafeteria to how every prisoner looked every day since entering prison. Usually, it would suffice to compare someone entering a cafeteria to recent data about all of the prisoners (for example, comparing the weight of a prisoner walking into a cafteria to recent weights of all prisoners in that prison).
However, consider a slim person who entered prison, gained weight over five years, and escaped five months ago. His weight may be moving back toward his pre-prison slimness. The biometric system should consider all of his weights (or at least may of his weights) while in prison, not just his weight on one day. The biometric system remembers weight and many other traits for years. The biometric system could compare the face of someone passing an airport checkpoint to the face of each former prisoner: on his first day in prison, his last day, the middle day of his time in prison, and fifteen randomly chosen days he was in prison. Maybe a prisoner's face was tanned in summer and pale in winter. The system should have data describing every variation of his skin darkness (or at least many variations). Years of detailed, identifying data are an immense advantage in trying to recognizing someone.
An advantage of this system is that it has much experience recognizing the subject. The system recognized the subject a few times a day, perhaps several times a day, when he was in prison (for example, when he passed through a checkpoint in a cafeteria, a sleeping room, a holding cell, an infirmary, a work area, or a recreation area). Later, the same computer system (or a different system with the same identification software and data) is asked to recognize the same person as he passes through a checkpoint in a bus terminal, subway station, athletic stadium, or other place.
Recall that the system could be used to recognize prisoners (for example, to count them) and to recognize a subject who had been a prisoner. When the system is used to try to recognize a subject who had been a prisoner, there might be variation in the amount of information the system stores for that purpose about each subject. If he's urgently sought, there might be much information about him in the system (for example, information about him on 200 days). Otherwise, there might be less information about him in the system (for example, information about him on only 20 days). In general, the more information the system has about a subject, the more likely he is to be corectly recognized by the system. For example, facial recognition is much worse when based on only one photograph than on multiple photographs.
Above, we briefly describe a simple system. One might have fancy versions. For example, different people customarily walk at different speeds. A biometric system might notice how quickly prisoners walk. Speed of walking is different from stride length, which is mentioned above.
Many identification traits vary over time. For example, a prisoner might exercise outdoors much in warm weather. As a result, he might be tanned and relatively low-weight from May to September, and pale and relatively high-weight from November to March. In a simple system, one might simply compare an observed person's weight to weights of people while they were in prison. In a fancy system, one might compare a person's August weight to weights of people while they were in prison in August. One might experiment with many traits to see how useful they are in automatically, biometrically recognizing prisoners and wanted ex-prisoners. We think that the system would be an excellent way to count prisoners, track them as they move within a prison, and recognize them after they become free. Many of the people sought by law enforcement agencies are former prisoners.
A strength of this system is the variety and amount of information the system has about each person whom it tries to automatically recognize. However, the system could also try to recognize subjects using little information (for example, try to recognize a subject using only one height, one weight, and facial structure information from only one photograph). The system would make more mistakes when trying to recognize those low-information subjects.
Implementation of the system would be in a few steps. First, one implements the system in a prison. In a prison, prisoners are automatically identified as they go through a checkpoint. The system can regognize every prisoner. Then, one implements the system outside prison; for example, at an airport. Finally perhaps, one mounts the system on trucks which can be parked almost anywhere; for example, on a street. As people walk along a sidewalk, they are automatically identified (for example: face, colors, height, stride length, speed of walking). This would be a portable system. A portable system probably would not use weight as one of the identifiers.
The biometric system we describe above does not consist of parts every one of which already is highly dependable. A few of the possible identifiers may not be available yet.
Many of the examples above invovle prisoners. This is for two reasons: it's convenient to get identifying information about prisoners (they're in prison), and the biometric system (which learns from experience) identifies each prisoner many times. Before a prisoner leaves prison, the biometric system will be able to recognize him.
In principle, the system could be used to recognize anyone for whom the necessary information (for example, photographs) is available, not just prisoners or former prisoners.
There are two steps: getting information about a subject into a biometric system, recognizing people who match that information. Getting information: the biometric system we describe above gets information about somone because the staff puts the information (for example, photographs of the system) into the system. Recognizing people: the system recognizes people about whom it already has information. The staff decides whether to put information about someone into the system. It would be possible to have a system which would get information by itself. Consider a biometric system checkpoint in a hallway at an airport. The system occasionally recognizes people passing through the checkpoint. One day, a man walks through the hallway. The system does not recognize him, so it saves the information about him that it got as he walked through the checkpoint (for example: numbers which are derived from photographs the sytem took of him as he went through the checkpoint, height, weight, length of stride, speed of walking). The system assigns a name to him (Unknown Subject 123, or US123). He has now been entered into the system. He is a subject who was not entered by the staff. When a person walks through the checkpoint later, the system tries first to match the person to subjects the staff entered. If there is no match, the system tries to match that second person to the Unknown Subject people (in other words, to people whom the system entered into itself as they walked through the checkpoint). If the second person matches a staff-entered subject or US123, fine. Otherwise, the second person becomes US124. The system saves information about him: identifying information and information about when he passes through the checkpoint. If the same biometric system is in a railroad station, the system might recognize US124 going through a checkpoint in the station. If a biometric system were installed in many places, it might recognize many people as they go hither and yon. The system described above might be useful to transportation planners, who often are curious about, for example, how many people who go through a specific place in a railroad station go through there at least: once daily, once every business day, once weekly, once monthly, and so forth.
The system would automatically search for duplicates. For example, when a new prisoner's information is entered into the system, the system would check if he closely resembled other prisoners. Sometimes, two prisoners happen to resemble each other. Sometimes, the same prisoner enters prison twice, each time under a different name. The system could easily tell the staff about prisoners, if any, whom the new prisoner closely resembles.
Consider a building in which employees wear identification cards. Machines throughout the building observe and recognize the cards. The machines are connected to a central computer. The computer knows that card number 12345, issued to employee John Smith, was not in the building yet today. The machine knows that card 23456, issued to employee Jane Jones, was in the eigth floor hallway outside room 810 at 9:15 a.m. today.
The computer knows that card 34567, issued to employee James Brown, was in the seventh floor hallway outside room 720 at 10:25 a.m. today (fifteen seconds ago). The computer is almost certain tha the person wearing the card was not Brown. The wearer was an inch shorter, weighed twenty pounds more, had lighter-colored skin than Brown, had a bald spot, and walked faster than Brown walks. Brown usually smokes when he's on the seventh floor, the computer has noticed. The card-wearer was not smoking. Furthermore, the card-wearer's traits do not match any employee except maybe Jerry Jenkins. The computer notifies a security officer that someone other than James Brown, maybe Jerry Jenkins, was observed wearing Brown's card in the seventh floor hallway outside room 720 fifteen seconds ago. The computer shows the security offficer a video clip of the card-wearer and color photos of Brown. The computer automatically shows information about Brown and information about the card-wearer (for example, height and weight of the two people) with important differences in red. After comparing the information about the card-wearer to the information about Brown, the security officer decides how to handle the situation.
This paladium.net Web site and all of its contents are provided as is, solely for informational purposes, and without warranty. There is no express warranty, no implied warranty, and no warranty of fitness for an intended purpose. We are not liable for informational errors or incompleteness. We have no duty to update anything. Verify all statements and claims. Carefully read and consider many sources of information, analysis, and opinion. Please inform us of each and every falsehood in this Web site. We own all email messsages and other communications to us. This Web site contains parody and humor. Our linking to a Web site is not approval or endorsement. This Web site, which is not exclusively about government officers, identifies some people who never were government officers. Our discussions and editorials do not necessarily apply to anyone named or identified in this Web site.
Home Addresses of People Most of Whom Are Current or Former Government Officers
| |||
|---|---|---|---|
| publicity about this site | search this site | an idea | research tips |
| world's governments main page | previous page | next page | |
