Acoustic Echo Cancellation with Algorithm

Abstract

Speech is the most basic form of human communication. Still, to convey this information, noise interference in voice signals and waves the most common problem in speech processing must be addressed. Sounds, Equipment, traffic, people, and echoes are familiar auditory sources of interference in speech processing. The project report addresses the problem of cancelling out undesired speech that is communicated through a microphone. In essence, echoes are produced when a feedback loop is started. This feedback loop starts when a microphone detects the near-end speech that a speaker makes when communicating. For instance, when we talk, the microphone detects our voice and sends it to a distant place again, causing the speech at the other end to echo back. Reverberation or the sound waves reflected off all the surfaces in a room is another echo source. When speaking, when something is produced in a room, the microphone there records the direct speech from the source (person or speaker) as well as the numerous reflections created by the architecture of the space, which arrives slightly later than the direct speech and causes a delay (echo). Together, these reflections create a single continuous sound. Half-duplex communication is only one way to solve this issue; in this method, the near-end's microphone is switched off whenever the large-end speaks, preventing the short-end from interrupting until the far-end has finished speaking has ended their discourse (which was one-sided). An echo cancellation is the primary fix for this issue. When simultaneous speech communication is required during audio teleconferencing, echo cancellation is crucial. This study discusses echo cancellation, a system comprising adaptive filters and algorithms. Several parameter tests were conducted to determine the best algorithm for cancelling echo.