Voice Processing


Automatic Gain Control
Often an audio signal firsts needs to be amplified before it can be processed because the distance from the microphone to the audio source can vary across a large range. Therefore KWest has developed a configurable AGC module. All parameters, such as Reference Output Amplitude, Maximum Output Amplitude, Gain Raise and Fall Times etc., can be set during run-time.

Voice Activity Detection
KWest offers two solutions for Voice Activity Detection: an LMS-filter-based solution to differentiate stationary voice from voice signals, and a Frequency-Domain-based solution that analyzes spectral changes within the range of the human voice to detect voice activity.

Noise Reduction
There are many solutions for Noise reduction. Selection of the most appropriate solution depends on the available hardware and the acoustic environment. KWest can offer Time Domain-based and Frequency Domain-based solutions with Single- or Multi-Microphone input for Stationary or Dynamic Noise Cancellation.

Echo Cancellation
As for Noise reduction, there are different options for selecting the most suitable approach for the given hardware. KWest can offer Echo Cancellation based on LMS or MDF algorithms in Time or Frequency Domain. The Echo Tail Length corresponds to the Filter Tap Length and can be freely configured during run-time.

Vocoder
For digital voice transmission and recording, a variety of different codecs exists to compress speech signals. KWest can offer simple ADPCM-Codes, GSM Codecs, and Speex, e.g. for VoIP.