Audio identification isn't dead .

This board is open for discussion on the Australian Yowie, Bigfoot, Yeti & Sasquatch. Please keep on topic in this forum.
Post Reply
User avatar
Rusty2
Long Time Contributor
Posts: 1396
Joined: Mon Aug 23, 2010 5:30 pm
Position: Believer
Location: East Coast

Audio identification isn't dead .

Unread post by Rusty2 » Wed Jan 10, 2018 9:18 am

I got to thinking this yesterday morning about the charts and realised I'm using a 2 dimensional chart to represent subjects and impacts without taking into account distance which means I need a 3 dimensional chart to calculate 1. impact frequency , 2. distance and 3. multiple subjects .

So , now I have 3 wallabies on video and recorded correctly by the audio recorders .
I went back to those places yesterday , found the exact place where the cameras caught the wallabies and measured the distance to where each wallaby was located .

Earlier in the day I went over my recent walk test and calculated my average impact at every 2 metres up to 30 metres . At a distance of 30 metres my footsteps are barely audible and going any further would be pointless if we can't see or hear them .

I added all the figures to a 3d Excell chart and found something interesting . The wallabies are topping out the chart at just 14 metres whereas my impacts take a lot longer to rise because I have a heavier impact .
The grey columns in the back row are our wallabies . Yes , I undertsand that there will be difference in wallaby weight but the figures in the chart are correct and represent 3 wallabies at 3 measured distances . I'll also be adding more wallabies and roos to the chart as I collect them , so far so good .

The orange columns are my measurements at every 2 metres . You can see a great difference between the two .

The lone blue column is only a stab in the dark at the moment just to show what it may look like .

So how do we determine the weight of an unknown subject ? We don't have to . What we have to determine is distance of the unknown subject . If we can determine distance of an unknown subject then determining it's weight may be irrelevant .

So how do we determine the distance of a unknown subject on an audio recording ? With a fence . If I put up a cheap fence around the audio recorders to prevent anything approaching any further than a maximum of 10 or 15 meters then we'll know it's outside a particular distance .

Black cotton thread could also be used outside the fence at various distances to determine the amount of intrusion . This may not help if we collect more than one visitor .

Any thoughts ? Anyone see a problem ?
You do not have the required permissions to view the files attached to this post.

User avatar
ChrisV
Long Time Contributor
Posts: 623
Joined: Mon Nov 10, 2014 3:28 pm

Re: Audio identification isn't dead .

Unread post by ChrisV » Wed Jan 10, 2018 11:57 am

Talking about black cotton - the one thing that gets me every time is when the thread is broken at 7-8 feet high! That is so fascinating because there is essentially nothing with the exception of bird flying or a possum breaking it which is unlikely.

Not sure that helps your case but its a tangible thing the 'ol cotton test!

User avatar
hillbilly
Gold Status - Frequent Poster
Posts: 255
Joined: Wed Jun 01, 2005 11:35 am
Position: Believer
Location: Blue Mountains

Re: Audio identification isn't dead .

Unread post by hillbilly » Wed Jan 10, 2018 2:36 pm

Should the cotton be trusted? The cotton could be affected by birds, both day and night. Also bats, or even falling branches could impact the cotton, at any height. A large kangaroo in full flight mode might reach 7+ foot, also impacting on cotton. Climbing reptiles could impact the cotton where it may be tied off, at any height.
The fence? Would the fence allow access for a rabbit? When hunting a rabbit area I have heard rabbit "thumps" from 25 metres away. If that rabbit was next to your recorder there would be a very loud thump. Would we assume it to be something large?
Wallaby bounding impact sounds would vary according to their situation and fear. A scared wallaby makes very loud thuds as it escapes.
You did say "Any thoughts?" :)

User avatar
AustralopithecineOz
Long Time Contributor
Posts: 573
Joined: Sat Oct 01, 2011 11:16 am
Position: Nature Lover
Location: Riverena NSW
Contact:

Re: Audio identification isn't dead .

Unread post by AustralopithecineOz » Wed Jan 10, 2018 5:39 pm

Excellent Rusty.. when you have refined the method I would love to see the results from my procession of subjects recording. It still baffles me. Good work mate!

User avatar
Rusty2
Long Time Contributor
Posts: 1396
Joined: Mon Aug 23, 2010 5:30 pm
Position: Believer
Location: East Coast

Re: Audio identification isn't dead .

Unread post by Rusty2 » Wed Jan 10, 2018 6:31 pm

Good points with the cotton , looks like it's ruled out , thank you .

Hillbilly , yes I've heard rabbits thumping too , quite a noise . I can't say I've ever seen any rabbit activity out in the mountains where I go . I don't usually go near any farms for fear of losing my gear .

From what I remember , rabbits thump but that's about it yeah , no walking type sounds , someone correct me if I'm wrong . Yep , frightened wallabies and roos may make louder sounds . I do know a place where roos bolt down this ridge at dusk to get to the green grass at the bottom of the hill . If I was to collect some audio and video from that area then we could add those impacts and step lengths to the charts .

Dave , I'm not sure if I can save any recordings from the past because we don't know the distance of the subjects , but I do have a plan to see if I can . Remember that drop test I did years ago ? It failed because I couldn't calculate the weight of the impacts . Now I can . there's a mob who make pressure sensitive film . The film shows the weight of the impact in coloured ink .
I still have all those measurements and may find something interesting but don't hold your breathe , it may not be possible .

User avatar
AustralopithecineOz
Long Time Contributor
Posts: 573
Joined: Sat Oct 01, 2011 11:16 am
Position: Nature Lover
Location: Riverena NSW
Contact:

Re: Audio identification isn't dead .

Unread post by AustralopithecineOz » Wed Jan 10, 2018 8:36 pm

All good Rusty.. I am embarking on another round of recording this week and it should go on for some months yet. Hopefully I can get some more unusual fodder for you to tinker with.. well thats the plan anyway. (thumb up)

User avatar
Rusty2
Long Time Contributor
Posts: 1396
Joined: Mon Aug 23, 2010 5:30 pm
Position: Believer
Location: East Coast

Re: Audio identification isn't dead .

Unread post by Rusty2 » Thu Jan 11, 2018 7:53 am

Hey Dave , if you can't determine the distance of any recorded footsteps , I may not be able to help you at all .

Although cotton thread seems to be inconclusive it may be cheap insurance for you if it were kept low , like waist height . If upon your return you find that the thread hasn't been broken it would at the very least provide us with a known distance .

Hope you get something cool !

User avatar
AustralopithecineOz
Long Time Contributor
Posts: 573
Joined: Sat Oct 01, 2011 11:16 am
Position: Nature Lover
Location: Riverena NSW
Contact:

Re: Audio identification isn't dead .

Unread post by AustralopithecineOz » Thu Jan 11, 2018 8:10 am

Thanks Rusty.. I'll give it a go.. Anything that might help.

User avatar
Slats
Silver Status
Posts: 106
Joined: Wed Mar 08, 2017 7:48 pm
Position: Field Researcher
Facebook Profile Page: https://www.facebook.com/WestOzYowieResearch/
Location: Perth
Contact:

Re: Audio identification isn't dead .

Unread post by Slats » Thu Jan 11, 2018 11:19 pm

Hi Rusty
Wouldn't your audio radar provide a better way to obtain distance to subject? It maybe possible to extrapolate the distance to subject by a ratio between recorded volumes and your baseline data footsteps.

Maybe????

User avatar
Rusty2
Long Time Contributor
Posts: 1396
Joined: Mon Aug 23, 2010 5:30 pm
Position: Believer
Location: East Coast

Re: Audio identification isn't dead .

Unread post by Rusty2 » Fri Jan 12, 2018 8:35 am

Hey Slats , that radar was only for vocals , woodknocks and foot thumping . I could calculate which direction they're coming from but haven't figured out a way to clearly define distance .
There are noticable differences in some footsteps audio which I know about and that you can see for yourself but I havent looked at nutting it out yet . It seems to have a lot to do with weight . The lighter the subject , the more indistinguishable it seems to be .

Below are screenshots of a 10 kilo weight dropped at a height of 1 metre at a distance of 12 and 39 metres . The 12 metre signature is a full line . The 39 metre signature is missing all the mid frequencies , thats the difference but lighter footsteps aren't as clear as this .

The unknown footsteps screenshot shows the same thing , missing mid frequencies , but how far away is it ? I can't tell .
You do not have the required permissions to view the files attached to this post.

User avatar
Slats
Silver Status
Posts: 106
Joined: Wed Mar 08, 2017 7:48 pm
Position: Field Researcher
Facebook Profile Page: https://www.facebook.com/WestOzYowieResearch/
Location: Perth
Contact:

Re: Audio identification isn't dead .

Unread post by Slats » Fri Jan 12, 2018 10:40 am

Caution this may be confusing!

Firstly, I think the use of kangaroos and wallabies in the data set is somewhat pointless. It can be used to rule out subjects but comparison to yowie/human bipedal walking, in my opinion, is useless as the mechanics of movement is completely different and therefore could not be used in any comparison.

Secondly, Are you able to see volume in your recorded data? If so the audio radar should be able to be used to calculate distance by comparison of volume.
Say, for example, your steps at 8m=20dB, 16m=10dB and 32m 5dB. You should be able to draw a graph showing a curve.
Comparison of the mass of a subject should be done with similar data i.e. you weigh 100kg and I weigh 140kg, volume of our steps at the same distance would be comparable data.
You could then increase the mass of a subject by using a backpack containing a set mass and conducting the same tests volume at the extreme limit could be tested and then you would have the outer limit of your radar.
These tests carried out at various distances and masses would give you a baseline data set to compare with.

I think it would be a reasonably safe assumption that only adult yowies would approach your recording station in the initial set up within a location as it would appear as a potential threat.
Then by using direction of movement in conjuction with volume a distance could be calculated along with an estimate of the subjects mass.
This approach may eliminate the use of a fence/cotton trip lines that may injure or impede other wildlife.

User avatar
Slats
Silver Status
Posts: 106
Joined: Wed Mar 08, 2017 7:48 pm
Position: Field Researcher
Facebook Profile Page: https://www.facebook.com/WestOzYowieResearch/
Location: Perth
Contact:

Re: Audio identification isn't dead .

Unread post by Slats » Fri Jan 12, 2018 11:30 am

Furthermore, if an average adult male gorilla weighs 200kg that would mean the average weight per foot of height is 33.33kg per ft. The average yowie hieght is 6ft10 (sightings with hieght estimate source "The Yowie" Tony Healy and Paul Cropper) that would make the average weight 212.75Kg.
You could use the data set to calculate distance by comparison of volume and weight.

Also, frequency can be used calculate distance.
Lower freqencies travel further. This is represented in your screen shots, the further the distance the lower the frequency that will be recorded. So at your extreme limit of recordable data it will only be at certain frequencies.
I believe distance could be calculated with the average freqency recorded of each step.

Austral
New Member
Posts: 39
Joined: Tue Nov 28, 2017 7:10 am

Re: Audio identification isn't dead .

Unread post by Austral » Fri Jan 12, 2018 6:54 pm

What about some type of seismic probe to measure the ground disturbance?

gregvalentine
Gold Status - Frequent Poster
Posts: 263
Joined: Fri Sep 23, 2016 7:42 pm
Position: Crypto Enthusiast
Location: Sydney

Re: Audio identification isn't dead .

Unread post by gregvalentine » Fri Jan 12, 2018 10:42 pm

Austral wrote:What about some type of seismic probe to measure the ground disturbance?
Probes should perhaps be left to those pesky aliens . . .

User avatar
Rusty2
Long Time Contributor
Posts: 1396
Joined: Mon Aug 23, 2010 5:30 pm
Position: Believer
Location: East Coast

Re: Audio identification isn't dead .

Unread post by Rusty2 » Sat Jan 13, 2018 10:17 am

Great points Slats , let me have a look at it and get my brain around it .

Hey , Austral , probes sound expensive , Greg might be right .

User avatar
Scarts
Gold Status - Frequent Poster
Posts: 352
Joined: Sat Aug 31, 2013 7:33 am
Position: Researcher

Re: Audio identification isn't dead .

Unread post by Scarts » Sat Jan 13, 2018 11:52 am

Infra sound detection with sound recording, in conjunction with ultraviolet cameras or filters, and thermal cameras.

Austral
New Member
Posts: 39
Joined: Tue Nov 28, 2017 7:10 am

Re: Audio identification isn't dead .

Unread post by Austral » Sat Jan 13, 2018 3:04 pm

gregvalentine wrote:
Austral wrote:What about some type of seismic probe to measure the ground disturbance?
Probes should perhaps be left to those pesky aliens . . .

Yeah well, i thought it was a good idea.

User avatar
Rusty2
Long Time Contributor
Posts: 1396
Joined: Mon Aug 23, 2010 5:30 pm
Position: Believer
Location: East Coast

Re: Audio identification isn't dead .

Unread post by Rusty2 » Sun Jan 14, 2018 9:39 am

I have been busy but wacked these figures out this morning . The table shows volume versus distance from myself and wallabies . Although the chart isn't finnished the interesting part is that it's reasonably close . The numbers in the chart are actually negative numbers but using negative numbers turns the chart upside down and adds confusion . The lowest columns are the loudest impacts . There's a lot of work to do to get this method up and running but it seems plausable and I'll be looking into it .

If only I had the time , money and patience Scarts . Are you doing anything in this area like you mentioned a year or 2 ago ie , hyperspectral imaging ?

I had a look at your suggestion Austral about seismic probes , they do make them but no one is giving away any prices to little fish like me which means they're probably very expensive . There's lotsa intrusion detection devices http://drivewayalert.co.nz/
including microwave and buried underground devices http://www.smartwatchcentral.com/Securi ... tdoor.html
The driveway alert with single beam seems to be the go but setting up and aligning it may not be worth the trouble or money . I'm not even sure if the base reciever is battery operated .
You do not have the required permissions to view the files attached to this post.

User avatar
Slats
Silver Status
Posts: 106
Joined: Wed Mar 08, 2017 7:48 pm
Position: Field Researcher
Facebook Profile Page: https://www.facebook.com/WestOzYowieResearch/
Location: Perth
Contact:

Re: Audio identification isn't dead .

Unread post by Slats » Thu Jan 18, 2018 10:12 am

Hey Rusty it looks promising so far and seems to be more useable data. I think if you can draw a curve rather than bars once you get some more data it will be easier to get the distance in later recordings.
You could then have one axis of the graph showing negative through positive decibels and the other axis distance. I would estimate it would look something like the photo.
You do not have the required permissions to view the files attached to this post.

User avatar
Rusty2
Long Time Contributor
Posts: 1396
Joined: Mon Aug 23, 2010 5:30 pm
Position: Believer
Location: East Coast

Re: Audio identification isn't dead .

Unread post by Rusty2 » Thu Jan 18, 2018 4:17 pm

Hey Slats , the above chart (subject volume versus distance) is actually what you talking about but at the moment it's so short we can't see a curve . Below is the same chart with added calculations , just changed the name to footsteps volume . You can see my data is rising waves . Zero (0) is the loudest volume . It used to be 100 but someone changed it for some reason , now it's back to front . ?

I may have to use the chart below to estimate distance then use the very first chart at the top of the page (subject impacts versus distance) to then show distance versus impact or use both so we can see two views . Two views to show something out of the ordinary .

Maybe then we can estimate a weight but that will be a while . I'll have to redo the footstep audio in the dark to maximise sensitivity of the recorder .
Thanks for contributing to this Slats !
You do not have the required permissions to view the files attached to this post.

User avatar
Slats
Silver Status
Posts: 106
Joined: Wed Mar 08, 2017 7:48 pm
Position: Field Researcher
Facebook Profile Page: https://www.facebook.com/WestOzYowieResearch/
Location: Perth
Contact:

Re: Audio identification isn't dead .

Unread post by Slats » Thu Jan 18, 2018 4:31 pm

All good mate anytime. I look forward to seeing the results

Post Reply