Max Lewis, Roshin Roychan, Speech Recognition for Access.pptx

zy

1. Speech Recognition for Password-Based Authentication Authors: Roshin Roychan, Max Lewis CPSC 4155 10/29/22

2. DEFINITION Speech recognition: AKA automatic speech recognition (ASR), computer speech recognition, speech to text, etcetera Broadly defined capability enabling programs to process human speech into written format Commonly confused with voice recognition Interesting fact: IBM released “Shoebox” in 1962; could recognize 16 words (IBM, 2020)

3. Key Concepts Password authentication offers easy way of verifying user access Post Covid era innovation drives contactless environment By combining password authentication with speech recognition; access points such as doors can have second layer of security or contactless entrance

4. Issues No access to Google’s source code; only access to Python’s library code Modifications are limited to length, modes, and cosmetic type features Formatting must be done post-translation Limited to free version (worse than paid?) Must handle UnknownValueError Delay and syncing times

5. Scope Current focus is speech recognition for password authentication Voice recognition would require much machine learning Simplicity for learning curve Few desired outcomes Multiple interfaces unnecessary SCOPE

6. Status/On-going Research Current implementation works with simple phrases and words Future features may include formatting for more complex passwords Considering adding complexity of machine learning and voice recognition

7. Breadboard Wires LED lights Resistors raspberry pi USB mic GPIO breakout board ribbon cable Items Needed

8. Connect USB mic to port connect ribbon cable to GPIO port on raspberry pi (colored cable closer to SD card side) Connect ribbon cable to GPIO breakout board Connect GPIO breakout board to bread board Ensure negatives line up with bread board and GPIO; All GPIO pins should nest on bread board Connect green wire to 17, yellow wire to 27, and red wire to 22 Connect LED lights green, yellow, and red with longer legs facing away from GPIO Connect resistors to negative on bread board and in same row as short legs of LEDs Connect wires to their respective color LEDs Assembly

9. Type alsamixer in raspberry terminal Press F6 to choose sound card Mic Check Press F6 to choose sound card

10. Press F5 and turn down mic threshold Further confirm by listing capture hardware devices by typing arecord -l Mic Check(cont.) Type lsusb to list USB devices connected Confirm your mic is listed as device See what sound card your mic device is on Type command below to change default config file Change defaults.ctl.card and defaults.pcm.card to sound card number your device is on

11. Reboot system to set default Mic Check(cont.) Play recent recording back Test mic with recording a sample; use command below to facilitate

12. Check Python version and if it is lower than 3.92 than use following command to update and install Python Instructions Install pyaudio both ways Install Python speech recognition library Install Free Lossless Audio Codec for Google API Further install some FLAC dependencies Ensure pyaudio installed websocket associations

13. Use both commands below to install wiringpi Enables the use of the GPIO breakout board Wiringpi maybe an included library in Python Python Instructions (cont.) Time to code and demonstrate

14. Python Code and Text

15. Conclusions Speech recognition is great for authentication when considering convenience of contactless effort Speech recognition has advanced to a point of being accurate to a high degree of confidence Uses for security and access are trumped by other methods stemming form the same genre (voice recognition) The advantage to using speech recognition for access is that it is easy and requires no machine learning. Good for public scenarios where notice of user is unknown. As a stripped-down transcriber, it is unmatched Speech recognition may still be viable security option pending further research

16. References AssemblyAI. (2022). Build a Speech Recognition System on a Raspberry Pi . YouTube. Retrieved October 30, 2022, from https://www.youtube.com/watch?v=vEMzN5RgXbw. BMonster Laboratory. (2022). Make A Raspberry Pi Guessing Game In Thonny Python IDE Using A GPIO Breakout Board And One RGB LED. YouTube. Retrieved October 30, 2022, from https://www.youtube.com/watch?v=6Kw7TnIl5hY. Electronoobs. (22AD). Voice Recognition Raspberry Pi and Arduino Uart communication . Electronoobs . YouTube. Retrieved October 30, 2022, from https://www.youtube.com/watch?v=TMHVLVSsVTs. IBM. (2020, September 9). Speech Recognition. IBM Cloud Learn Hub. Retrieved October 29, 2022, from https://www.ibm.com/cloud/learn/speech-recognition