This project delves into various aspects of audio signal processing, specifically focusing on echo creation, cancellation, and background noise classification. Each part of the project addresses distinct challenges and techniques in audio processing.
In the first part, two echoes are created by delaying the audio signal and adding gain. These echoes are then fed back into the original signal, simulating acoustic environments. The final output is normalized to the range [-1, 1] to ensure consistent amplitude, prevent clipping, and improve perceptual quality.
To cancel the echo, the signal is first autocorrelated to identify the peaks corresponding to the echoes. Custom envelope detection functions are used to accurately detect these peaks. Once the peaks and their indices are determined, their attenuation factor is calculated. A filter function is then applied to remove the echoes from the signal, preserving the original audio.
In the third part, background noise classification is performed without removing the noise. The audio signal is cross-correlated with four test audio samples representing different types of noise: ceiling fan, pressure cooker, water pump, and city traffic. A threshold-based approach is used to detect the presence of each type of noise in the audio recording.
This project showcases the application of signal processing techniques in various audio processing tasks. From creating and canceling echoes to identifying background noise, the project demonstrates the effectiveness of innovative algorithms and methodologies in audio signal processing.