PSTN Authentication

$1 is far too much, but a penny would do the job well I think. It would end the robocalls I think.

Of course you just don't take CID blocked calls. I think they have that already as part of the phone service.

--

Rick

For some reason I've not seen that on the cellphone, just the landline.

I actually got a spoofed CID call the other day that was MY OWN number

...Jim Thompson

--
| James E.Thompson                                 |    mens     | 
| Analog Innovations                               |     et      | 
| Analog/Mixed-Signal ASIC's and Discrete Systems  |    manus    | 
| San Tan Valley, AZ 85142     Skype: skypeanalog  |             | 
| Voice:(480)460-2350  Fax: Available upon request |  Brass Rat  | 
| E-mail Icon at http://www.analog-innovations.com |    1962     | 
              
I love to cook with wine.     Sometimes I even put it in the food.

Nope. What you're really trying to accomplish is block junk calls. Some kind of authentication system is just one of many possible solutions.

Reading between the lines, it looks like you're heading for a computer based automatic attendent or possibly and IVR (interactive voice response) system. You could probably purchase an IVR system and do most of what you're proposing. I've setup three for various businesses in the past. They work reasonably well and will block most attempts to circumvent the system. Basically, disarm the default "Punch zero for operator" function and you'll be fine. Please don't get clever and have the caller do a math problem and punch in the result on the touch tone pad. That has already been proven to be a bad idea because of the condition of the American educational system. Such IVR systems are boring and offer little in the way of entertainment value for the product designer. They also require some equally boring programming and configuration in order to operate properly.

Instead, I propose detecting the callers accent, and route the call to either the desired party, or to a black hole depending on the their accent. There's plenty of research on the speech recognition and accents. For example:

I suspect such accent filters might already be available. I've received telemarketing calls from what I presume is India, that started out with a pre-recorded message in perfect English, but when I started speaking, was taken over by someone with a very thick and unintelligible accent. The only reason I could think of doing that is to bypass an accent filter. Therefore, some IVR style filtering will be required, such as asking "Who is the president of the US of A"? or "What time is it"?

Yet another idea. Many years ago, I had a problem on the local ham on our radio club VHF repeater system. We had a legally licensed user that was becoming a nusance. He was asked to not use the radio club repeater, but refused. I decided to build a filter, set to detect this operator, and have the repeater turn itself off when he appeared. I tried using various radio signature identification schemes, which proved not sufficiently reliable. Next, I devised a speech recognition system that compared a recording of him identifying with his call sign, with the current digitized receiver audio stream. If it detected a match, the repeater would turn off for a few minutes. It worked amazingly well, until he found out what was happening and changed his manner of identification.

It might be possible to arrange for a recording of anyone that has an excuse to call your phone and compare the recordings with the same phrase spoken by the caller. In other words, an audio password. Saying their own name should be sufficient to identify and authenticate the caller. This has the advantage of not requiring someone on a smartphone to punch buttons while driving and using a hands free system, where the keypad is not easily accessible.

Good luck with your project.

--
Jeff Liebermann     jeffl@cruzio.com 
150 Felker St #D    http://www.LearnByDestroying.com 
Santa Cruz CA 95060 http://802.11junk.com 
Skype: JeffLiebermann     AE6KS    831-336-2558

I had an IR motion detector wired to my computer and media player. When I woke up in the morning, it would turn on these devices. When I left for the office or went to sleep, they would turn off after a 15 min delay. I had some falsing problems cause by reflections from passing cars, neighbors borrowing tools, post-midnight fridge raids, and my time share cat, but in general it worked well enough to determine if I was home. An IR motion detector room occupancy sensor could be built into the authentication system to detect when one is asleep.

If this ultimate telemarketing call filter just happens to be a computah, the answering machine function could be built in. A common

8GB SDHC card should be sufficient for recording messages and storing a call record. The rest is software.

--
Jeff Liebermann     jeffl@cruzio.com 
150 Felker St #D    http://www.LearnByDestroying.com 
Santa Cruz CA 95060 http://802.11junk.com 
Skype: JeffLiebermann     AE6KS    831-336-2558

There ya go... you're racist... you will be visited shortly by the thought control police ;-)

Seriously, that's an excellent idea!

[snip]

Those Indian call centers drive me nuts. ...Jim Thompson

--
| James E.Thompson                                 |    mens     | 
| Analog Innovations                               |     et      | 
| Analog/Mixed-Signal ASIC's and Discrete Systems  |    manus    | 
| San Tan Valley, AZ 85142     Skype: skypeanalog  |             | 
| Voice:(480)460-2350  Fax: Available upon request |  Brass Rat  | 
| E-mail Icon at http://www.analog-innovations.com |    1962     | 
              
I love to cook with wine.     Sometimes I even put it in the food.

Practice typing "computer". Your use of computah is somewhat annoying. Or, is that your goal?

No. What I really want is an "electronic secretary".

"I'm expecting a call from Tom. Please find me when he calls!"

"If Bob calls, tell him I'm on my way!" (as I head out the door)

If a client calls while I am asleep (I have a wacky sleep-wake cycle), take a message.

If SWMBO calls, UNCONDITIONALLY find me (it may be important)

If calls "too early" or "too late", tell him/her: "Is this

*really* important enough for me to disturb him?? Or, would you rather I just tell him that you called??"

The "secretary" analogy/model is a good one, I feel. A "good" secretary knows your habits, preferences, priorities, etc. and can make decisions based on those.

E.g., if you are "in the conference room", chances are, you won't want to take a call from a friend/acquaintance/sales rep. If you're in "the corner office" -- along with several other executives -- an incoming call from your SO should probably have a higher threshold set for interruption: "Should I interrupt his meeting with the execs?" ("No, just tell him I called and ask him to please get back to me when he can")

What do we do with friends, neighbors, clients, etc. that have the "wrong" accent? :>

Access to services like that are best not *gated*. It is too obvious. (like software that refuses to run because it is "unlicensed").

A more effective deterent is to *degrade* the service so it just looks flakey/buggy. E.g., inject "noise", "randomly" drop the carrier, etc. So, the unwanted user doesn't realize he is being targeted but, instead, thinks the service is "crap". (and walks away INSTEAD of trying to figure out how you are blocking him)

I think I should be able to characterize their speech - given enough of a sample. E.g., we all have "accents" of differing degrees. How many folks mispronounce "salmon"? Or, say "yes" as "yeah" or "yup"? "The" as "thuh" or "thee"? etc. Plus, the coloration of their vowels, etc.

Getting the "reference sample" should be easy -- just listen in on a conversation with that party. Analyze it when you have "spare MIPS". Then, update a database with those characterizations (along with the CID encountered, time of day, day of week, etc. -- anything from which you may be able to deduce a recognizable pattern!)

The trick, thereafter, is to be able to map your future encounters with this (as yet unidentified) individual with the database of previously encountered callers. And, do so interactively (so an adversary can't just play a recording of a trusted individual saying "hello")

*Toys*! :>

My apologies for only being somewhat annoying.

I have no plans to adjust my spelling to accommodate your personal preferences. However, as a consolation, I'll try to use the dictionary spelling instead of the way it's commonly pronounced in our correspondence and usenet replies. Feel free to remind me should I forget.

No. If I wanted to annoy you, you would have been visited by at least

5 out of 10 biblical plagues, various government officials trying to be helpful, and telemarketers with incomprehensible accents. The electrolytics in your electronics would all simultaneously overheat, outgas, and leak. Everything you own would refuse to continue functioning without an expensive upgrade. All your power devices would be recalled as fire hazards. Your monitor would be stuck in 640x480 mode. You would discover that your friends are all simulations and live in virtual machines. I would also have constructed a clone of your computah, err... computer, stick pins into the motherboard, and by the magic of sympathetic voodoo, your computer would fail in the same manner. Since none of the aforementioned disasters has happened to you, it would be fair to assume that I have no interest in annoying you.

Well, maybe a tiny annoyance. Google reports 59,400 hits for "computah". Seems to be growing in popularity. About 2 years ago, it was half that number.

Also, thanks for the link to Smith v3.10. It looks quite good and I'll give it a try.

--
Jeff Liebermann     jeffl@cruzio.com 
150 Felker St #D    http://www.LearnByDestroying.com 
Santa Cruz CA 95060 http://802.11junk.com 
Skype: JeffLiebermann     AE6KS    831-336-2558

A neighbor has a "security system/service". PIr detectors in every room with an outside "exposure" (i.e., means by which an intruder could potentially gain access -- windows, doors). Thus, essentially *every* room has one! (except the guest bathroom which is "land-locked" -- and the "hallway")

One could envision other technologies used to detect occupancy, activity, *identity*.

[A local hospital track the locations of every employee on the campus]

Just knowing that an individual is in a particular room allows a lot to be inferred (with some degree of accuracy) about his/her activities.

In a residential setting, there are only a few things that *tend* to happen in a bathroom (shit, shower, shave, etc.). And, the amount of time spent in the room can further refine the likely activity deduction. If you can observe other things (e.g., are the lights on/off? exhaust fan on/off? water running? etc.) then these observations can add to your confidence re: a particular "diagnosis".

Exhaust fan on, water running at a significant rate --> shower. Exhaust fan off, water running at a modest rate --> washing/shaving. Exhaust fan on, occupied for a long time, no water --> smelly sh*t! :> etc.

Of course, you can make a *wrong* deduction (maybe the user is repairing some of the plumbing and has the fan on to vent the solder/flux fumes from the propane torch)?

The same sort of things apply to institutional environments -- the conference room *tends* to be used for meetings (if you can track employee A, you can also know that employees B, Q and T are *also* in that same room! More likely having a meeting than "playing poker"!)

[Note residential environments are harder to instrument -- people are far more sensitive to "their money" than businesses are (stockholders' money!). And, most businesses PLAN on a certain amount of "facilities remodeling" in their annual budgets. Also, commercial properties tend to lend themselves more readily to "modification" than do residences.

Famous last words: "the rest is software"! :>

It's been tried before:

It's much easier to have the electronic secretary run your life for you than for it to guess what you're going to do next. For example, the electronic secretary could prepare your schedule for you and enforce compliance with the schedule through various unpleasant ways. I vaguely recall a 1950's sci-fi story that revolved around that theme.

The problem is that you're building a state machine with an unknown number of states. You can do that successfully, but only to the point where you go insane trying to define all possible states for all possible input conditions. There are AI (artificial intelligence) programs that will do that. I've scribbled one program each in Prolog and LISP to see what they were like. More successfully, I worked on a telephone answering contrivance, that used a state machine to determine the outcome of all possible input conditions. The state table was wall size. (The flow chart fit on letter size page). It's not like you have only two possible states for your machine (i.e. answer phone, don't answer phone). You also have a wide collection of possible actions that can be performed for either state. What makes this work is that with a state machine, everything happens at the same time with storage for prior states. It can be done with a big lookup table if necessary.

There are various schemes to help simplify the programming. One is to assign a "value" to each input and a "cost" to each output. For example, on a scale of 0-9, a call from the family would be a 9, while an obvious telemarketer detection would be a zero. Various acquaintances would have assigned "values" as would your location, time of day, and what else you might be doing. Any state that exceeds some threshold, answers the call. Below that threshold, the call is dealt with according to a "cost". For example, dumping to voice mail is cheap. SMS message alert to your cell phone is expensive.

The down side to such a state machine is that you can easily create undefined states. That's where a combination of inputs has not been defined and the state machine has no clue what to do with it. An early ATM machine did that when someone forgot to program in the input state of the cash drawer. The machine thought the transaction hadn't been completed and failed to deduct the amount from the users account. Repeating the cash withdrawal over and over until the drawer was finally opened, was recorded as a single withdrawal.

Your electronic secretary has far too many input states to be manageable. Off the top of my thinning head:

1. CID (caller ID)
2. Time of day
3. Local activity (meeting, lunch, busy, working, etc)
4. Your location
5. Remote activity (driving, meeting, travel, airport, bus, etc)
6. Urgency level
7. Human telemarketer detection
8. Machine telemarketer detection
9. Value of caller
10. Cost of call disposition (Voicemail, SMS, ring through, forward)
11. Occupancy detector (did you step out?)
12. Alarms and alerts (computer failure, heartbeat, autoreboot, connectivity loss, power fail).
13. Priority bypass and over-ride.
14. Remote message playback request.
15. Local intercom.
16. Whatever else I forgot. Now, all you need to do it put together a chart for every possible condition and state of these inputs, to decide what to do with the incoming phone call. Got the picture?

Lots of ways to bypass the system:

1. CID (caller ID) from known caller. Note that you should check both the calling number and the associated ID. Both can be spoofed, but it's unlikely that a telemarketer will get the ID correct.
2. Touch tone a bypass code.
3. Voice recognition bypass.
4. Hangup and call back. System detects two sequential calls from the same CID.

Actually, the trick is to have more than one recording of each individual with which to compare. Unfortunately, that's not always possible. Time for some math. Ideally, it would be someone saying the same word or phrase through various devices (POTS, VoIP, cell phone, Bluetooth). Each sounds sufficiently different to make identification possible. I had the advantage of a phase that was present in most transmissions (the FCC call sign) which also provided a form of identification. Very roughly, my address book has about 500 names. If I store 5 seconds of a phrase, with 6 different devices, I'll have 30 seconds of recordings to compare per caller. With 500 possible callers, that's searching 250 minutes of recordings per call. I gave up trying to compare compressed audio, so that's about 2 Mbytes/minute or 500 MBytes of data that has to be searched and compared continuously for each call. If I'm looking at a 5 second window, then I'll be searching at: 500 MBytes * 8 bits/byte / 5 sec = 800 MHz I other words, my brute force method is not going to happen with a commodity PC.

--
Jeff Liebermann     jeffl@cruzio.com 
150 Felker St #D    http://www.LearnByDestroying.com 
Santa Cruz CA 95060 http://802.11junk.com 
Skype: JeffLiebermann     AE6KS    831-336-2558

I worked on a proposal for a nuclear power plant that wanted to know the location of literally everyone at all times. We didn't get the contract, but did participate in the inevitable compromises until the bitter end. The problem was that they wanted to know if anyone was trapped inside a room or building in the even of a nuclear incident. There were various systems proposed for identification (mine was an IR transponder on an ID badge) but all were rejected as being unusable during an emergency, especially when everyone tries to squeeze through the doorway bottlenecks. After the usual acrimonious yelling and screaming, it was decided to keep it simple and just track people going in and out of the doorways. There were only 20 doorways that needed monitoring. They gave up on individual identification and just settled on counting people going in and out which was done with two PIR detectors per doorway. They couldn't do activity and identity, but occupancy was possible.

Expanding on this, you could issue everyone an RF ID badge, and track their location. The technology for this already exists. There are some technical problems, but the real problem is human. People forget their ID badges. The local hospital has such a system. I can get real numbers if you want, but my guess is that on any given day, at least 5% of the staff has forgotten their electronic door keys, ID badges, or both. They pickup temporary cards at the security office and then forget to return them.

True, if you're consistent. I'm not. For example, I arrive anywhere between 8AM and 1PM. lunch is anywhere between 10AM and 3PM. I leave anywhere between 5PM and midnight.

I'm sure the family will be thrilled with your requirement that they carry RF ID cards and have their activities monitored.

I've worked on several "smart home" systems. Something like this, but more customized: One of them has a voice control feature. You say the magic words and the computah, err... computer, responds with the desired action. The system has some intelligence. For example, telling it to turn on the lights when the lights are already turned on, will produce a computerized question asking for clarification. There are sensors scattered around the house to deal with the HVAC system which includes an occupancy sensor system. It's not to track users so that phone calls can be directed to the correct room. It's simply to turn off the heat or cooling in rooms that are not being used. Batting average for getting right is about 95%, which means that the HVAC system is wasting expensive gas and electricity for 18 days per year. Swell.

Automation is fine. Predictive automation is a problem (see DARPA autonomous vehicle trials for an example).

Have you considered that this project might be a bit too ambitious? If not, have you considered that instrumenting the house and office in the manner you suggest might be overkill for simply answering a telephone? Rube Goldberg comes to mind, where the complexity of the system is far in excess of what is required to accomplish a simple task.

That works for me because I'm not a programmist.

--
Jeff Liebermann     jeffl@cruzio.com 
150 Felker St #D    http://www.LearnByDestroying.com 
Santa Cruz CA 95060 http://802.11junk.com 
Skype: JeffLiebermann     AE6KS    831-336-2558

If you "dial" while a call is in progress, you stand a good chance of dropping the call. (At least this is how things *used* to be. I can dig out a dial phone and try it on our current service. Of course, no guarantee that that will speak for the nation as a whole...)

"Easier" implies "easier for the DEVELOPER". I prefer to makelife easier for the *user*!

VRT/IVR systems are only *increasing* in popularity as vendors try to replace minimum wage humans with zero-wage machines.

If you try to address an *unconstrained* vocabulary, the problem is far more difficult than if you inherently limit the user to picking from a shorter list of words (which can be *deliberately* chosen to be more easily "distinguishable").

For example, the selections {beam, beat, bean, beak} would be a

*really* bad choice! OTOH, {beam, sick, tan, case} would be much more reliable (note I have just made up two four-word sets without concern for the *application*)

This is why it is relatively easy to RELIABLY recognize things like account "numbers" in speech. And, why resllving SPELLED words would be considerably harder (more choices -- 26 vs 10 -- and more "close" sounds -- 'ess' vs. 'eff', 'bee' vs. 'pee', etc.)

How does making coffee relate to "PSTN Authentication"?

See above.

Secretaries used to answer the phone and "take messages". Then, we created answering machines! Secretaries used to "take dictation"; now we have Dragon Dictate.

Folks who previously couldn't "afford a secretary" *can* afford an answering machine, speech recognition software, etc.

You aren't thinking hard enough! :> There are lots of ways to determine *without* GPS! When you go shopping in your favorite department store, the overhead cameras aren't just watching to have a record of any shoplifting activities; rather, they are also tracking your movement through the store and, eventually to a cash register (and a specific *transaction* at that cash register!).

If you stopped in "Sporting Goods", then, presumably, you were

*interested* in those products (you don't go looking for bath towels in that department!).

If you're in the kitchen, you are probably either:

- preparing a meal

- cleaning up from a meal

- routine housecleaning

- grabbing a quick snack/beverage etc.

You probably are NOT:

- taking a crap

- sleeping

- reading a suspense novel

- mowing the lawn

- changing the oil in your vehicle etc.

"Place" is a pretty good predictor of "activity". You can further qualify this with simple things like "time of day" (e.g., chances are, if I am in the kitchen at 5P then I am probably doing something "meal related"), accumulated knowledge of the user's habits, etc. (I have a friend who "sleep eats"!) Note that you dont care what

*specifically* the user is doing in those places. Rather, you are concerned with the *types* of activities in which he *may* be engaged!

If you're in the kitchen (for any length of time -- ruling out "grabbing a quick snack/beverage), then the activities in which you are likely to be involved will *tend* to be similar -- in terms of your willingness to "answer the phone", "respond to an incoming email", etc.

In a business environment, equivalent patterns exist: the Monday afternoon staff meeting; hosting sales reps just after lunch; debugging in the lab; etc.

Who said "using a door"? C'mon, I'm disappointed at how unimaginative you're being! You *can't* come up with an ECONOMICAL way of telling where I am in a particular "edifice"??

The industry trend seems to be contrary to your intuition -- witness all the money moving into home automation, etc. Apple wants to "have your house welcome you home" (by sensing when your cell phone returns to the premises). You don't think they haven't considered watching your cell phone's "travels" through the house to *also* cater to you?? (how many folks effectively *wear* their cell phones? Almost like a piece of JEWELRY! Like, maybe, a WATCH!!!)

Because that requires the user to "do everything". What good is a secretary that has to be told how to handle every contingency? A *good* secretary learns your preferences, habits, priorities, etc. and adapts them to new situations.

E.g., if you chose to NEVER take a call from Fred "after 9PM", then why can't it learn from OBSERVING your behavior (like your secretary) and *hide* incoming calls from Fred at those times? Why should the user have to EXPLICITLY "program" every action? (i.e., "told how to handle every contingency")

Connecting your smart phone to a bunch of BT peripherals (lights, HVAC, doorbell, etc.) does *nothing* (that you couldn't also do in

1970) to change your way of life. It just saves you a few steps to the lightswitch, thermostat, etc.

What you want from "systems" is for them to anticipate your needs and adapt to them.

E.g. instead of "programming" your thermostat to different temperature settings at various times on various days of the week, why not just

*change* the temperature to be what you want it to be -- and let the thermostat *watch* what you have been doing (over the course of days, weeks, months, etc.). You end up with a very different UX!

"Gee, with my smartphone, I can call home and turn the ACbrrr on

*before* I get home so it is comfortable when I arrive!"

(Um, why doesn't your ACbrrr know to turn *itself* on in anticipation of your returning home at 5:13-5:17 eh evening??)

See above.

We are creature of habit. At home. At work. etc.

If you have a device that is *on* 24 hours a day, 7 days a week, 365 days a year doing nothing *but* answering the phone (and watching

*hen* YOU choose to do so, etc)then why can't it LEARN from those observations instead of forcing the user to *tell* it how to handle each condition?

If you carefully look at how you live *your* life, you'll find there are big patterns to which you have grown so accustomed that they are invisible!

What time do you go to bed? Wake up? Do you shower just before bed or just after rising? Coffee in the morning? Or, only on those mornings when you "weren't late"? Late night snacks? Read before bed? Leisurely showers or "functional" ones? Stopo at the gym on tuesdays and thursdays?

Because it doesn't have the capabilities that a "good secretary" would have! And, it is too tedious for you to have to tell it how to handle every contingency!

AND, something that can SEE how I respond to EVERY situation can make some deductions from that! (coupled with some general rules that I can pre-state to "prime the pump").

This is how a good secretary works. He/she "knows" you.

I want more than that. I want to know who is calling so I can use that information to *tailor* a response.

E.g., if I am "expecting a call", then I really DON'T want most of the normal rules to apply! I want to "keep the line free" (because, even with Call Waiting, I can't know the identity of the *waiting* caller!) waiting for *that* caller.

If *I* call in, I may want to "leave instructions" for how calls should be handled in the immediate future ("If Bob calls wondering where I am, tell him I've had car troubles and will be a little late"). Or, retrieve messages.

The BIGGER problem is: "This is Dr Welby calling from The Local Hospital. Your offspring/spouse is here in our ER in a life threatening situation. We need you to come in and give your consent..."

("Telemarketers, please press 1. Friends and family, press 2. Doctors looking for emergency contact, press 7" :-/ )

How do you handle the folks who YOU may want to have access to you (via the phone) but that may not feel obliged to deal with your "system"?

"Screw this! Let him *wonder* why his kids haven't returned from school, yet. Or, hear about the accident on the evening news..."

I've already got those "poor" solutions. And, no sign that ANYONE is "thinking about it" -- AT ALL (let alone "forever"!)

C'mon...this isn't a "new" problem. And, doesn't rely on "new" technology. So, why are all ofthese solutions that "do the job" leaving *so* much to be desired? (why do you *still* receive telemarketing calls? political surveys? etc. "No market" for such a GOOD device???)

I guess they must think they are doing a DIFFERENT job!

You keep making jumps in logic. An IR detector can't tell if you are asleep or not. It can't tell if you are at the office.

It also can't tell if you have gone to the kitchen or bathroom or whatever. It is very hard for a computer system to tell what you are actually doing and why you are in a given room. I know in my lifestyle this sort of thing would be pointless.

Yep if that is what you want. But that doesn't get you a cordless phone. Mine already has the answering machine built in complete with remote access. Why reinvent the wheel?

--

Rick

Incidentally (an intro to topic drift), you can get the same effect using Google Voice Search. It should work if your computah, errr... computer, has an attached microphone. Run Google Chrome. Go to: Settings -> Advanced Settings -> Privacy and check the box labeled [x] Enable "OK Google" to start Voice Search Then, bring up Google search. You'll see a small microphone on the right side of the search box. Just say "OK Google" followed by whatever you want to lookup. If you have your speakers enabled, sometimes the answer will spew from the speakers.

I like to leave open a window with Chrome running so that I can ask stupid questions. Favorite questions are "what time is it" and "current weather". A favorite demo is "how far to Dominican hospital". It's also good for unit conversions. Try "convert 10 feet to meters" or "convert 22 miles per gallon to liters per kilometer" and the usual math problems. Also physical constants. Try "Boltzman's constant" or "Planck's constant". More:

It's not quite ready to run your life for you, but that will probably be a feature in the next release.

--
Jeff Liebermann     jeffl@cruzio.com 
150 Felker St #D    http://www.LearnByDestroying.com 
Santa Cruz CA 95060 http://802.11junk.com 
Skype: JeffLiebermann     AE6KS    831-336-2558

You must have been in some other country. I've never been able to drop a call by pulse dialing. In fact it used to be common for systems to accept pulses as well as DTMF.

Go for it! I won't wait for your solution. I may not live that long.

Yep, they work well when you are calling from your quiet living room. They are crap when calling from a cell phone on the highway.

Think

See above.

What does that have to do with *your* gadget? Will your gadget be a PC running Dragon Dictate?

So your system will tap into the store's video and look for you?

Or looking for the shoes I left somewhere other than the bedroom or kitting my kayaking gear or feeding the cats or looking out the window at the bird feeder or repairing my glasses with a steak knife or getting frisky with a lady friend... such a long list, my PC only has 16 GB, I'm not sure it can remember every possibility.

Other than the first two, I don't do any of those anywhere... Not very useful info to anyone.

I don't agree at all. Look at my list. Some are very interruptable and others.. well not so much.

Again, nonsense. I can be sitting at the exact same computer doing very different work that has very different levels of interruptability. Or not work at all again with different desired interrupt masks.

None of the above is any evidence that I am wrong. They make all sorts of useless gadgets that are sold all over the world.

"Learns"! Even a secretary has to be told.

Sounds great, but even my PC, arguably the smartest device I own, hasn't figured how to "learn" the simplest of my repetitive behaviors. You think my refrigerator will do any better?

Uh, my thermostat has already learned my behavior... for half the year

tell it when it is winter and when it is summer by changing the temperature on the thermostat and changing "cool" to "heat".

Because I work from home?

What have you been drinking? All of a sudden you are on an IA jag.

No, it is impossible for my phone to know the difference between being at lunch with someone I don't mind interrupting to take a call and someone I don't want to interrupt, etc...

I assume you don't mean "knows" in the Biblical sense?

And you have to *tell* the system that, right?

Again, this is far, far too complex to tell the system how to deal with Bob, Ted and Alice every half hour.

Yeah, your phone is going to recognize that message... lol

You are really out there.

We have discussed getting rid of 99% of the robo and marketing calls. There are even devices out there that do this. Most of what you suggest is simply not practical or won't work well.

--

Rick

Makes sense. Note, however, that they probably are happy to just know that there is *A* person in *A* particular room. The identity and EXACT location would be icing on the cake (the identity could be inferred as "one of the guys that are currently unaccounted" -- and probably refined beyond that: Which of these guys are likely to be

*in* that room?") [You also have to consider the mindset/motivation of the users -- in your scenario, one would assume they WANT to be found -- unlike Finney! :> ]

You can track presence a variety of different ways. Much depends on how many bodies you are trying to track and how many locations.

E.g., you can track RFID tags passing by fixed RFID readers (at "choke" points in the traffic flow). You have no idea where the people are n the times between 'reads'. But, geography can allow you to make good deductions (there's only one way into that room!)

OTOH, if you have relatively few bodies and many locations, you can affix the RFID reader (effectively) to the *body* and have it report tags that it "passes by"!

"Policy" has not incentivized them, suitably! :> "Your badge is your timecard. Without it, The System has no way of knowing that you worked, today!" (Draconian but you get my point)

This is the approach that I am currently pursuing. It serves two purposes: provides location tracking for the user(s) (by noting which device is "assigned" to each user) AND provides a means by which The User can interact with The System -- and vice versa.

It's the equivalent of carrying your cell phone with you "everywhere you go" (except IN the shower) but without the bulk/inconvenience/cost.

The "incentive" for doing so is you can interact with The System without having to wander over to some "control terminal". And, The System can better anticipate your needs (because it knows WHO you are, your preferences, current location and its impact on those preferences, etc.). E.g., if *I* call for "music", it will not likely be the same sort of material that SWMBO would want to hear!

I lead a very unstructured life. (SWMBO wonders how we ever *see* each other! :> ) Yet, I never shower in the kitchen. Nor eat in the bathroom. If I am in the bedroom for more than a few minutes (time it takes to restock the printer in there), I am asleep (or trying to be).

If I am in the Living Room, I am probably watching a movie. If I am in the back yard, I'm picking fruit or doing yard work.

Activities tend to stick to locations.

[One screw up is I cant resolve altitude. So, if I am working on the roof -- patching, painting, servicing antennae, etc. -- The System thinks I am inside the house (and wonders how I was able to move from the master bath DIRECTLY into the kitchen... through several walls!!! :< Unanticipated condition required some rework of my coding logic.]

How are you incentivizing them? Most folks that I know carry their cell phones on their person even inside their homes (Sheesh! Can't you WALK OVER to the counter to pick it up when it rings??) Would you complain if you wore a wristband that gave you access to The System (and, in return, tracked your location "for free")?

*My* preference is for a BT earpiece. It gives me voice access to The System so I can interact with it "hands/eyes free". And, I don't have to *build* the damn things (at least for protos).

If you have "data", then why limit it to one particular "application"? Why shouldn't the phone be able to benefit from knowledge of the user's location? Or, likely activities?

Knowing that the occupant(s) is(are) asleep allows you to adjust the HVAC. But, it also lets you inhibit incoming phone calls that could

*disturb* that sleep (i.e., if you can identify the caller and make a judgement call as to how important that interruption would likely be regarded by the sleeping occupant). Or inhibit the doorbell for similar reason.

Ensure the lights in the house have been turned off (one of my most common errors: "You left the lights on in the kitchen last night").

Etc.

Leverage *everything* that you have -- an "integrated" solution instead of a bunch of "islands of automation" (e.g., storebought solutions!)

You have to consider the consequences of a "bad prediction"! If someone's call goes to voicemail instead of to the callee, there may be some "opportunity cost". Or not.

If a car assumes you will ZIG and you, instead, ZAG... :<

Nope. I don't aim for a perfect solution. Rather, to put together a framework that others can improve upon (or, dissect and extract "useful bits").

My real goals lie elsewhere -- this is just a good vehicle to explore a "significant (understatement!) application" that touches on lots of different areas (not easily dismissed).

Phone is just one of many subsystems. But, rather than *simply* tying the phone in to The System, take the opportunity to improve upon how we interact with The Phone (and callers).

Relacing a secretary with a (tape) answering machine is a tiny improvement. Replacing a tape answering machine with a digital store is another tiny improvement. Ditto voice-mail (remote access, call forwarding, etc.) None of these things fundamentally change the way we interact with people trying to "access us" via the PSTN!

Google wants to peek inside your house. They don't want to rely on what you *tell* them (by the content of your emails, searches you conduct or sites you visit).

Rather, they want to see how you live your life so they can "help" you (by pitching other products and services to you! :> )

IMO, any tie-in to an "outside service" is just an offer to sell yourself -- often "for free"!

Notice TVs that "watch" the viewer(s) -- allegedly to allow them to control the TV without the need for a "real remote".

[Dear advertiser: the folks in this houshold were glued to the screen during that last commercial of yours! Perhaps you should send them some "unsolicited" coupons for your product... you stand a far better chance of them exercising those coupons than the household down the street -- where everyone left the room during your commercial!]

The same sort of things are being done with other staple appliances (refrigerators that watch what you consume -- to *help* you prepare a shopping list for this upcoming week... and, of course, tattle on your consumption patterns to boost sales for particular products, etc.).

I don't think it overly paranoid to want to exercise some control over "who sees what".

It doesn't have to guess what you are going to do next. Rather, know what the chances of your handling a particular situation in a particular way are likely to be.

E.g., if your "routine" is to listen to the news in the morning while rushing your teeth, then, chances are, if you command the "RADIO" on in the morning while you are in the bathroom, (i.e., no, you don't want the radio to be on in the living room or kitchen -- because you aren't LOCATED in those places!) it is LIKELY that you will want it tuned to the news station (and not the Jazz station that you were listening to in bed the night before).

If it errs, then the user is inconvenienced but not harmed. (i.e., you now have to call for NEWS instead of just the radio/TV)

No. You don't treat it as an FSM. Instead, you recognize "likely situations" (morning, bathroom, user1) and the likely actions that will you will be called on to perform in each of those situations.

E.g., (radio,news) is more likely -- by looking at the users historical behavior -- than (radio,music). Every further command issued by the user tweaks those probabilities. How you weight these changes is dictated by how quickly you want the system to "learn" them.

[You can walk in the bathroom tomorrow and command the RADIO and be met with NEWS or MUSIC... depending on how quickly you've configured the system to adapt to the change in your preferences the day before]

You don't deal with all inputs. You create simple rules ("invariants") and let it adjust the probabilities based on observation.

E.g., I *know* SWMBO should beable to get in touch with me REGARDLESS of everything else (in the shower, asleep, in the back yard, working on the car, out buying groceries, etc.). Rather than letting a system learn these things over a *lifetime* (how often will she call while you are inthe shower?? how long will it take to learn that you want to receive her call in that UNLIKELY situation?), "prime" the system with known facts.

You dont store recordings. Instead, you exract parameters from speech samples that you have *casually* observed (by listening in to the caller's side of the conversation EVERY TIME THE USER SPEAKS WITH HIM.

You make estimates of the various vowel frequencies, etc. Then, tabulate these.

You notice how particular words are spoken (you have lots of time between phone calls to analyze all this recorded speech!). E.g., my regional accent causes some embedded 't's to bespoken as a glottal stop -- effectively cutting the word in half at that point. I pronounce 'g's very hard. Others effectively *omit* that sound in many words (esp "ing" -- as "in'")

Now, the "real time" (interactive) problem becomes one of steering thecaller into saying words from which you can easily extract these parameters -- and, doing so in a casual/natural manner.

Just follow the bouncing ball.

I do sleep in my office, so that might be a problem. For excursions outside, to get some air, I had a long time delay before it would turn off the computer and medial players. I think it was about 10 minutes. Note that I was more interested in whether I was active and in the house than whether I was in which room. I had 3 motion sensors upstairs and 2 downstairs. The logic was that if ANY of these showed activity, I would be considered NOT asleep.

The basic assumption is that if NONE of the motion sensors detected any motion, I would be considered either out of the house, asleep, or dead. A simple pressure switch on my bed, couch, and overstuffed chair would be sufficient to determine if I was asleep. Maybe a microphone to detect snoring would be helpful. The lack of any senor input would show that I was gone.

It's not pretty, 100% accurate, or perfect, but methinks good enough.

A motion sensor in the kitchen and bathroom would take care of that. If there's no motion in the computer room, but motion is detected in the kitchen or bathroom, I would have added a few minutes before the power dropped off on the computers and media players. Again, it's not perfect logic, but I think it's good enough.

Creeping featuritis? The electronic secretary has to interface with the telephone system in some manner. Might as well have it do everything involving the phone including the cordless phone aspect. Even the ancient Merlin phone system that I had the pleasure of ripping out of the adjoining office had a built in 900 MHz analog cordless phone as an option.

Because options tend to be more profitable. Reinvented features such as the cordless phone option tend to be high profit margin items. The idea is to give away the basic functions and soak the customer for the accessories and options.

However, the soaking is not always because of greed. It's often because handling options are expensive. It's sometimes cheaper to just build them into the basic product and include them with every device, than to deal with the separate handling, packaging, documentation, inventory, etc. Worst case nightmare is with mutually exclusive options, where you can pick this option, or that option, but not both at the same time. I had that experience in the past and never want to repeat it. Integrating the answering machine, call logger, cordless phone, VoIP, Asterisk switch, T1 PRI expansion, baby monitor, and kitchen sink, might be desirable if handling these as separate options or as interconnect problems are deemed overly complex or expensive. Instead of reinventing, think of it as repackaging.

In this case, building in the answering machine makes sense because the box already has audio processing, disk storage, and CPU horsepower sufficient to easily add an answering machine. The IVR functions of the box closely resemble and answering machine. Most of the cost is software, which only needs to be bought once. On the other foot, the cordless phone is quite separate from the answering machine. Just because most cordless phone vendors decided to conglomerate both devices in a single box doesn't mean that such a combination is mandatory. Because the location of the box and the cordless phone base radio might not be the same, it makes sense to add the cordless phone bases outside the box and connected with phone cable.

--
Jeff Liebermann     jeffl@cruzio.com 
150 Felker St #D    http://www.LearnByDestroying.com 
Santa Cruz CA 95060 http://802.11junk.com 
Skype: JeffLiebermann     AE6KS    831-336-2558