MOTIVATION

In many places just taking a simple walk can be a dangerous activity due to violence, harassment or other acts that may lead to injuries, accidents or even traumas. One specific group of people that suffers due to not safe streets are women: many of them feels unsafe and often must deal with situations where they feel that they are being followed.

One possible approach to make street walkers safe is to somehow send a message to someone, generally calling for help, or taking another action such as cross the street, walk/run faster, or fetch a protective device (i.e. taser, pepper spray). From the technology point of view, there are many apps and wearable devices that provide a panic button functionality with many options to alert or notify someone of the situation with important data, such as GPS coordinates. However, current “panic button” solutions are activated only after the person that feels the threat suspects or know what is happening, which sometimes is too late.

INTERVIEWS

To gather more information about this situation and understand better the point of view of someone who is being followed in the street (or have the feeling of being followed) I did some interviews with a couple of women, including my wife. Unfortunately, this situation is more common that I was aware of, which makes me a little bit sad because of what a lot of women went through. This kind of situation make me eager to think about something that I could do.

First, I asked if the interviewed had ever of being followed or feel that was being followed. Many of the ladies that I talked confirmed that it happens a lot, mostly on streets and at night. The ladies mentioned that in these scenarios it is common to give a fast look over the shoulder to take a sneak peek at what is happening behind them and try to spot someone that is approaching from behind of even following them from a distance. Furthermore, they reported that they try to be discrete in their actions to avoid the follower to note their suspicions. Also, the stories they told are full of fear and panic if they found that the follower realize they are a possible victim. Another common attitude is to not use the mobile (fear to show any valuable item that could raise the interest for a mug) or any other gesture that signals their apprehension.

DESIGN

Based on the interviews I started to think of a possible design solution to help someone that feels to be followed in a suspicion’s way. I also remembered that the 2021 Hackaday prize had a category for hardware design solutions that involve displays. Therefore, I started thinking about the design of the solution that somehow involves displays and can help the in the “felling of being followed” situation.

During the initial design I thought about some spy movies where the main protagonist has the feeling of being followed. In these scenes it is common to see that the spy uses a mirror, a storefront glass, or any other reflective surface to look behind and note that some bad guys are looking and following him/her. We also have similar scenes in movies were cops are trying to do unsuspicious surveillance, most of the time inside a car eating donuts and drinking coffee.

One of the references that came to my mind was the rear camera that a some of cars have. They provide a helpful visualization when parking and some of them have proximity sensors and virtual bars to help the driver see the boundaries of the parking space fitted to the car size.

After some research I did some sketches and come up with a solution: a rear view camera that the women would wear as a wearable to help see from behind and be aware of a possible suspect following her. Here are some requirements that I gathered from the interviews:

These are general requirements and I know that I could not implement most of them due to the limitations of the time/resources I have available to work in this project. However, here I explain how I would approach them.

For the processing  I  choosed a Raspberry Pi Zero W with the official case (https://www.raspberrypi.org/products/raspberry-pi-zero-case/). This kit has a nice opening to host the camera, which can be a normal or night vision (https://www.raspberrypi.org/blog/build-your-own-raspberry-pi-night-vision-camera/). Also, there are some small portable LiPO batteries that could be used. I choose the RPI ZeroW because of its small form factor and computer power to run any kind of computer vision algorithm to do facial detection, record the video and even use GPS receiver. I could go with a EPS32 CAM or similar, but I don’t think they have a good form factor and processing power as the Raspberry Pi Zero.

The idea is to connect the RPI to mobile using WiFi to send the video stream to a display. However, I also thought about other display approaches that could be more discrete such as a smart watch, smart glass or a display like the Vufine VUF-110 Wearable Display (https://www.amazon.com/Vufine-006011-Wearable-Display/dp/B01MZ89QXF) that could be attached to a glass.

Since the displays available are not that great to show a full HD video that allow a better view of the back I designed a solution that will allow the person to interact with the video. The interaction will be to move a virtual “window” that shows a cropped (viewport) part of the video to fir better the resolution of the display (and also provideo zoom in the video ). In my design the user could use a power point clicker around his finger to control the position and zoom of the window that will fit the size of the screen (the smart watch or the a smart glass). I really liked the ring form factor of the AMERTEER Wireless Presenter (https://www.amazon.com/AMERTEER-Presenter-Presentation-PowerPoint-Rechargeable/dp/B06Y5S3JHL/ref=sr_1_8?dchild=1&keywords=logitech+clicker+finger&qid=1623333984&sr=8-8) that has four buttons, a USB dongle to connect to the RPI zero and a very discrete ring finger format that could be used without anyone that behind you notice. Another option to resize the viewport of the video is to use the touchscreen interaction of the smart watch.

The RPI zero W + camera + battery must be placed with the camera faced behind the user in a very discrete way. My initial though was to disguise the RPI case with a hair clip, hair pin, scrunchie or a barrette in the format of a flower to conceal the RPI but still allow the camera to see. This device could be placed in the woman hair or in other convenient places such as a backpack or the strap of a bag facing backwards.

There are several machine learning algorithms that allow face detection and my first choice was the Google’s Media Pipe face detection (https://google.github.io/mediapipe/solutions/face_detection.html ) because they have nice ports that runs just fine in for the RPI. Maybe some adaptation is needed to just track the face of someone that has appearing for some time in the video feed. I looked and did not found any of these face detection algorithms that provide some estimate of distance. Also, some of the existing face tracking algorithms can draw lines in the eyes to show were the preson is looking at, which may help the user of the system to know if someone is follwing the user.

.

Additionally, there are some interesting algorithms that provide information for blind people about the surroundings  including captions and other information about the surroundings that could be provided in the display.

EXISTING SOLUTIONS

The existing solutions for sake walking include a lot of mobile apps with panic buttons. There are also some physical panic buttons that are discrete, including bracelets, necklaces, and others. Those are great and can be used with the propose solution after the user note that someone is following her/him and there is a risk at some level.

The current solutions for rear view include the already mentioned car parking sensors kits , which includes cameras, small displays in the cockpit (even for the rear mirror near the celling of the car) and sometimes proximity sensors placed in the back rail of the car.

After some search I found another interesting scenario: street cyclists. They often need some sort of rear view and there is some commercially available mirror with rods. However, these are not discrete and practical for risk of being followed scenarios.

COMPONENTS

RPI ZeroW

RPI ZeroW Case

RPI ZeroW Camera

RPI GPS module

Battery Pack

Clicker Ring

A hack-friendly smartwatch

Hair clip/pin/scrunchie/barrette

LIMITATIONS

Unfortunately, due to the COVID-19 pandemic I was not able to get all the components I needed to assemble the project. I already got a RPI Zero, a camera, a battery pack, and the clicker ring. However, I was not able to get a display the live video, either a hackable smart glass (https://diyusthad.com/2020/03/top-4-hackable-smart-watch.html) or watch or a HMD (head mounted display) like the Vufine. I don’t know if I will be able to do that before the end of the 2021 Hackaday prize, but at least I have the entire design of the solution already documented. Therefore, the project is on hold for now.