Project
| # | Title | Team Members | TA | Documents | Sponsor |
|---|---|---|---|---|---|
| 62 | Voice Coded Lock |
Aman Thombre Logan Greuel |
Zicheng Ma | design_document1.pdf final_paper1.pdf photo1.png photo2.png presentation1.pdf proposal2.pdf video |
|
| # Voice Coded Lock Team members: - Aman Thombre (thombre3) - Logan Greuel (lgreuel2) ## Problem Currently, accessing secure areas usually requires some type of access card or keys, which can easily be misplaced, left at home, or stolen, leading to being locked out of your area of work or a security risk. These can also be hard to operate with your hands full - for example, trying to get an iCard out while accessing a lab in the ECEB while holding a laptop and FPGA can be quite difficult. Additionally, other keyless options requiring physical contact, such as a keypad, may pose a health concern for some users, especially during cold/flu seasons or a pandemic. ## Solution Our proposed solution is to implement an audio-based locking system for a door. We plan to create an attachment on or near a door which will listen for a user’s voice, and upon hearing a user saying some keyphrase, the device will automatically unlock the door. This system will use keyphrase recognition to identify an audio password, allowing for completely hands free operation of a keyless lock. ## Solution Components ### Subsystem 1: User Interface The user will interact with our locking system through a microphone and LEDs. Keyphrases will be listened for using a microphone, which will send the audio signal it records in real time to our microcontroller. An RGB LED will be used to signal to the user the state of the locking system - we plan to use different colors to indicate locked, unlocked, and listening. ### Subsystem 2: Keyphrase Recognition Audio signals will be fed from the microcontroller to a Raspberry Pi, which will perform keyphrase recognition. We plan to code this software in Python, and use some machine learning model (determined by which models give us best results in testing) to determine whether a given keyphrase is correct or incorrect. Signals will be fed from the Raspberry Pi to the microcontroller, signaling whether the keyphrase heard was correct or incorrect. ### Subsystem 3: Door Lock Operation We plan to use a deadbolt style lock which retracts when audio is verified. This can be achieved by using a motor and a screwing mechanism to engage/disengage the lock, or a more specialized moving component that can move side to side like a deadbolt lock. We also plan for this lock to retract in the event of loss of power as well, so that a loss of power does not lock users out permanently. ### Subsystem 4: Microcontroller on a PCB The microcontroller will be used to send signals to/from our user interface, perform some processing of the audio signal, and send signals to our door locking subsystem. The microcontroller will receive audio signals from the microphone, and when an audio signal passes over a certain threshold level, the microcontroller will send this audio signal to the Raspberry Pi for keyphrase recognition. When the audio signal is verified, the microcontroller will send a signal to the locking system to disengage the lock, and after a period of time, send a signal to re-engage the lock. The microcontroller will also send signals to our RGB LED, displaying whether the door is locked, unlocked, or whether a phrase is being processed. ### Subsystem 5: Power System Current electronic lock systems tend to use 4 AA batteries, so we plan to use up to 4 AA batteries to provide power to our microphone, LEDs, microcontroller, Raspberry Pi, and locking/unlocking operation. ## Criteria for Success - Keyphrase recognition should be able to consistently correctly classify audio password as correct/incorrect - Keyphrase recognition should be performed in reasonable time (<10 sec) - System should be able to operate a door lock automatically |
|||||