Chest Mechanisms, Wearables and Sensors

by fingerplayers

In November, I worked primarily on several chest mechanisms and the wearable. I also developed the overall structure and design of the puppet, the programming and the head structures. For this month’s reflection, I will focus on the iterations of chest mechanisms and share some of my work with the wearable and sensors.

Building the first “new” prototype

In my previous chest mechanisms prototypes, I faced the following issues:

  • The rise and fall movement of the chest were not obvious enough – the sponge chest was not expanding and contracting enough
  • Aluminium servo arms will cut into the sponge
  • Uni-directional rise and fall of the chest – My current sponge structure only expands and contracts from the front and all around the chest.

Moving ahead, I changed the material of the chest mechanism from sponge to EVA foam (see Figure 1). I examined various materials (ranging from sponge to boning structure and finally to EVA foam) and with some research, found that the rise and fall movement worked best using EVA foam installed in vertical strips. Figure 1 shows the process of building and experimentation leading up to the use of EVA foam. Video 1 shows the chest prototype using EVA foam. As a large piece of material by itself, EVA foam is unyielding and quite stiff as it cannot bend in the direction that I want it to. However, by cutting the EVA foam into strips, I could control the direction of the bends and achieve a “chest” that expands in all directions. I also added a boning structure which helps to hold the strips in place.

Figure 1: Process of building the structure for the new prototype using EVA foam.
Video 1: Chest structure with EVA foam that expands on all sides.

One issue I encountered when changing to EVA foam was the horizontal movement of the servo motors. Because the servo motors were moving horizontally and covered only a certain angle, the motor arms could only move one or two pieces of the EVA foam. I felt that I needed to find another solution that could achieve the effect I wanted without having to add more servo motors. Having more servo motors would add more weight for the puppeteer.

After much pondering, I decided to install the servo motors vertically. As such, the pulling action of the servo motors would be up and down instead of the horizontal in and out motion in the previous prototype. With this, not only did I not have to increase the number of servo motors used, I managed to reduce the original four to two servo motors.

To achieve these changes, I had to make many modifications to the old structure and rebuild it. Figure 2 shows my process.

Figure 2: Rebuilding the overall structure
Video 2: Testing the guide rods that the mechanism will move up and down on.

The Finger Players workshop is like Mary Poppins’ bag (or Doraemon’s magic front pouch, if you don’t know who Mary Poppins is). Although small, it has tools and materials that forms the answers to your problems. As I needed a part of the prototype to move up and down (seen in Video 2), I used the aluminium rods and wooden dowels found in the workshop to build this solution. Figures 2 and 3 show the building process to arrive at this step. Video 3 shows the simulated up and down movement of the completed top plate.

Video 3: Using my hands to simulate the up and down movement of the completed plate.
Figure 3: Building the guides and springs in the up and down mechanism.
Video 4: Testing the up and down plate with springs attached.

Subsequently, I hooked up the servo motors to the moving “platform” using steel wires between the servo motor arms and screw eyes in the “platform”. I modified the code for the servo motors (previously used to run the horizontal movement) to test out the up and down movement of the chest mechanism.

Video 5: First test of moving “platform” attached to the servo motors.
Figure 4: Close-up view of the chest mechanism.
Figure 5: Attaching the EVA foam to the chest mechanism.
Video 6: Chest mechanism with EVA foam attached.
Video 7: Trial of Chest mechanism with a piece of fabric placed over.

Refinements of prototype

I feel that the mechanism tested with the first prototype is promising in creating a chest that can expand in all directions. In my subsequent refinements, I will be:

  • Adding a hollow rod in the centre to run my wires
  • Adding an additional wooden piece at the top to build the neck joint
  • Shifting the positions of the servo motors to accommodate the centre hollow rod

Figure 6 shows the process of making the improved prototype of this iteration.

Figure 6: Building the next prototype to incorporate the refinements.
Figure 7: Interior structure of the chest mechanism.

Figure 8 shows my attempts to bend the wire rods at the neck joint and at the back support. As this was my first time doing it, I had trouble bending the metal rods which resulted in me not being able to complete the bends properly or making the hole too small for the puppeteer’s fingers.

Figure 8: Failed attempts at bending the metal rods.

I finally managed to bend the rods by not being too ambitious with the directions. Figure 9 shows the re-done neck joint and Figure 10 shows an enlarged base handle.

Figure 9: Neck joint metal wire.
Figure 10: Base handle metal wire.

Wearable & Sensors

After experimenting with both the flex sensor and accelerometer, I decided to omit the flex sensor for the wearable. This was because I felt that the accelerometer has more variations / permutations that I can play with and I wanted to focus on researching on the accelerometer to see how I can best optimise its functions. On the accelerometer, it can detect:

  • Activity
  • Inactivity
  • Single Tap
  • Double Tap
  • Free Fall

As its name suggests, most activity is detected when acceleration occurs. I am using the Sparkfun ADXL345 Accelerometer. I have been using the example codes from Sparkfun’s website at and tweaking the code to include the servo motor movements.

One of the difficulties I faced was the interruption of the various functions while using the accelerometer. For example: The servo motors are programmed to move at a certain speed and angle when no activity is detected on the sensor. When a fast movement is detected, I want the sensors to move quickly (e.g. move quickly 10 times), such that the movement of the chest mechanism is increased. However, the servo motors do not complete the 10 rounds. As soon as no activity is detected, the servo motors quickly revert back to its original speed (see video below)

Video 8: Servo motors being interrupted by the accelerometer input.

This would be the first problem I would like to fix. Besides this, I am also having difficulties triggering the “free fall” function on the accelerometer. In a more conventional robot set-up, the free fall function helps in detecting when the robot topples over the stairs or over the edge. How can this be triggered or simulated on a puppeteer?

I would like to have the accelerometer installed over the puppeteer’s shoulder as I feel that it is a position that would not be easily thrown off. The sensor will be secured to the puppeteer’s shoulder (non-master hand) so as to better differentiate between passive and active manipulation. If attached on the side of the master hand, I was afraid the sensor will be picking up too many readings caused by active manipulation that would override the readings of passive manipulation.  For the main structure of the wearable, I have decided to use a posture-corrector “back brace”. It is most secure and can be easily adjusted to fit various body sizes, as compared to the traffic vest that I was considering. I did a mock-up of the set-up in Figures 11 and 12.

Figure 11: Back view of wearable
Figure 12: Front view of wearable (connected to the puppet via wires on the puppeteer’s arm).

As I continue to prototype the wearable in the coming weeks, I will have to look into decreasing the number of connecting points and reduce the number of exposed wires to decrease the chances of any of the wires or connections breaking during performance. In my current mock-up, the jumper wires are connected at several different points and each connection is not well encased. Any movement from the puppeteer may cause parts of the connection to loosen or break due to the tension that the electronics may face.

In the next month, I will continue working towards a fully functioning puppet prototype. To achieve that, I will have to develop the puppet’s visual aesthetics alongside the technical hardware and programming software.

This article is a monthly reflection by Sim Xin Feng, the maker of our inaugural The Maker’s Lab as part of an ongoing 9-month experimental laboratory. The Maker’s Lab is curated and managed by Daniel Sim, a core team member of TFP. The ideas and reflections within the article are drawn from Xin Feng’s observations and discoveries as a maker, designer and researcher. Instead of being taken as conclusive, we hope that they serve to be a starting point for thought-provoking conversations and perhaps even debates. We would love to hear from you and can be reached at

If you wish to follow Xin Feng on her journey with The Maker’s Lab, click on the ‘Subscribe’ link below now!