Using hyper instruments that utilize sensors to extract gestural information for control parameters of digital audio effects is common practice. However, there are many pitfalls in creating sensor-based controller systems. Purchasing microcontrollers and certain sensors can be expensive. The massive tangle of wires interconnecting one unit to the next can get failure-prone. Things that can go wrong include: simple analog circuitry break down, or sensors wearing out right before a performance forcing musicians to carry a soldering iron along with their tuning fork. The biggest problem with hyper instruments, is that there usually is only one version, and the builder is the only one that can benefit from the data acquired. These problems have motivated our team to attempt to use sensor data so that sensors become obsolete. More specifically we use sensor data to train machine learning models, evaluate their performances and then use the trained acoustic-based models to replace the sensor.