Subscriber Discussion

Specialized Compression Hardware To Minimize Bandwidth Consumption?

CA
Chris Anderson
Mar 19, 2013

Hello

We work with a lot of federal agencies on covert surveillance projects, which ultimately end up using either cellular or microwave links. In each of these cases, bandwidth is limited. Cellular carriers are now going away from any "all you can eat" data plans, especially in the M2M (machine to machine) space, which this type of application falls under. As customers look to update to faster 4G connections they find that the carrier will no longer honor a grandfathered plan which once gave them unlimited data on the 3G network. As we all know, bandwidth isn't free and carriers cannot afford to allow single users to stream gigabytes of data, clogging their network. There are several things that can be done to look at addressing this situation, and edge recording is one that we employ extensively. That said, these types of customers still have requirement to access as much video as possible during their investigations. They also require solutions which work with commercially available cameras and carrier networks.

I am curious if the community has experience with compression hardware/software solutions that may fit. I have been exposed to very few in the past, which were hardware based client/server solutions such as Esstential Viewing (now Digital Barrier's TVI) offers - meant for pushing video over very low bandwidth GSM connections. I am certain there are military grade products that meet these demands, however I don't know if these have made thier way to commercial availability.

Does anyone have experience in this arena or recommendations?

Thanks

Chris

JH
John Honovich
Mar 19, 2013
IPVM

Chris, this might be a tough one.

Do you know how low you want to get for bandwidth per stream and what quality the user is willing to accept? For example, is it 500kb/s per stream, 100kb/s? They need 30fps, they are willing to go 5fps? 4CIF, CIF, etc.?

Also, the problem I've seen with many claims in this space is that it really amounts to simplistic transcoding to lower quality levels, which adds little value compared to just using a second stream, etc.

CA
Chris Anderson
Mar 19, 2013

Many of these guys are used to a CIF image at 5fps... Part of the challenge has also been that using H.264 either isn't acceptable or usable over cellular connections (things are somewhat better now with 4G bandwidth and latency). Also in law enforcement MJPEG has proven to be better for investigation and foresnics. This of course leads to much higher overall data usager per month when streamed 24x7, even at 5fps CIF images.

JH
John Honovich
Mar 19, 2013
IPVM

Chris, want do you want? MJPEG only?

Most of the specialized hardware compression I've seen uses a proprietary inter-frame codec which would likely have as many, if not more, issues than H.264 for the concerns your users have.

Also, why isn't H.264 usable over cellular connections? I can't see the technical reason for that.

CA
Chris Anderson
Mar 19, 2013

Sorry for any confusion. I am speaking to what has historically been used for reference. Due to the complexity of this problem customers will be open to looking at any and all compression technologies.

Regarding the experience we've had, especially with 3G connections when using H.264 is that the streams are unreliable. I believe this is due to the unpredictable nature of the 3G link, including latency and overall bandwidth availability. Visually we generally see these streams suffer from very poor performance which results in a very "jittery" or "bursty" live image. In many cases frames are dropped and the images become unreliable for investigative purposes. We have tried to adjust GOV as well as trying both CBR and VBR with no real tangible results. This is suprising to me as I believed we should be able to use a capped VBR to obtain good results - however our previous testing did not show this. That said it may be worth picking up again as we now have 4G networks available and have not done methodical testing yet. I agree that technically it should work great, but in practice we saw a different result.

Thank You,

Chris

JH
John Honovich
Mar 19, 2013
IPVM

Chris, are you saying MJPEG performed better? If so, what specific settings are you using for MJPEG? Also, what camera/encoder's H.264 were you using?

CA
Chris Anderson
Mar 19, 2013

John,

Since 'better' can somewhat be determined by use case, I can help clarify in terms of our customers application. MJPEG performed better in terms of reliability of the video, however it of course is limited in frame rate. So in terms of what law enforcement activities require, it is better to get fewer reliable frames than the jittery, bursty images we saw from the H.264 stream. Additionally as the streams are generally recorded, the MJPEG stream performs better in terms of forensic analysis.

The cameras we use are almost exclusively Axis and in this case it would be both the P55XX and Q60XX series of cameras.

JH
John Honovich
Mar 19, 2013
IPVM

Chris, knowing the camera manufacturer and type helps a lot (as implementations can vary).

I am assuming your users are remotely controlling these cameras, yes? Does that impact visible image quality?

Also, now that I know the manufacturer, I can ask Axis for technical pointers.

Avatar
Jeffrey Hinckley
Mar 19, 2013

I can understand the problems with H.264 if you were to stream video over cellular to a recording device. I have seen similar problems with remote internet (DSL, Cable) mesh networks, and other wireless PTP in noisy environments. You switch to JPEG, and problems go away (loss of frames does occur). I have not seen issues with Edge based recording though (Axis Companion, Bosch, Exacq Edge), since you only have to wait for the search buffer to fill.

Chris, can you clarify your setup and how the end user accesses the video? (Our Verizon 4G deployment in Maine is nearly completed). I am also interested in use rates you have seen (Since the Verizon plans are 10 GB limit).

I would also be curious of, depending on your method, the authentication characteristics for forensics if you are getting raw JPEG images for use.

New discussion

Ask questions and get answers to your physical security questions from IPVM team members and fellow subscribers.

Newest discussions