
7th ISF
PHYS018 - Personalized Human Speech Cancellation Using Synthesized Voice


This project is interested in developing the Active Noise Cancellation (ANC) technology. Previously, this technology could only cancel out sounds with low and constant frequencies. However, it struggled to cancel out human speech due to its complicated frequency components. Therefore, we aimed to advance this technology by eliminating vowel sounds since vowel sounds are essential components of human speech. If the vowel sounds are eliminated from the words, it can render those words incomprehensible. We scope our project by using only the /a/ sound. Our approach involved synthesizing the anti-phase sound waves of the target sound by repeatedly recording the vowel sound /a/. Then, we used the Fast Fourier Transform (FFT) to convert these recordings into the frequency domain, displaying magnitude values. We combined these frequency sets and phased them back to create the desirably synthetic sound waves. In the experiment, we placed two speakers facing each other, one emitting original sound and another emitting synthesized sound. A microphone was set up between them. We recorded the resulting sound and used FFT to analyze whether the specific frequency component had reduced and by how much. We also checked if other frequency components had increased. The preliminary results showed that the particular sound significantly reduced by approximately 17.75 decibels. In addition, we demonstrated the capability to cancel out sounds with multiple frequency components. Detailed research and refinement are progressing for more development and optimization for practical uses. We expect our work will contribute to the audio and entertainment industry and daily life. The existing headphones that are capable of cancelling only noise would potentially be able to cancel human speech as well. The dual benefit is reducing noise pollution. Moreover, in the livestock industry, it may help reduce the fluctuating frequencies that lead to animal stress, resulting in suboptimal productivity
Show MoreName :
Email :
Advisor :
School :
PROJECT QR CODE
