The MELPe algorithm was developed using several enhancements to the original MELP 2400 bps specification. MELPe implements a variable low bit rate vocoder that supports multiple rates of 600 bps, 1200 bps, and 2400 bps. It uses the MELP sampling rate of 8,000 kHz as well as the original frame size of 22.5 ms containing 180 digital voice samples.The nominal bandwidth ranging from 100 Hz – 3800 Hz is also used.
In addition to the 600 and 1200 bps rates, the MELPe algorithm also supports compressed bit-stream transcoding between the different rates. A new Noise Preprocessor has been implemented to reduce background noise while the original postfilter has been improved to further increase the quality of speech reproduction.
MELPe encoding is robust in difficult background noise environments such as those frequently encountered in commercial and military communication systems. The speech compression algorithm is very efficient in its computational requirements which translates into relatively low power consumption, an important consideration for portable systems.
- The MELPe Vocoder extends the traditional LPC parametric model with features for mixed-excitation, aperiodic pulses, pulse dispersion, and adaptive spectral enhancement.
- Mixed-excitation is used to simulate frequency dependent voicing strength with adaptive filtering and fixed filterbank. The multi-band mixed-excitation reduces the buzz associated with LPC vocoders, especially in broadband acoustic noise.
- MELPe vocoder generates synthesized speech using either periodic or aperiodic pulses. Aperiodic pulses are used to reproduce erratic glottal pulses without introducing tonal noises.
- Pulse dispersion uses fixed pulse dispersion filtering and a spectrally flattened triangle pulse to spread the excitation energy with a pitch period which, in turn, reduces the harsh quality of the synthetic speech.
- Adaptive spectral enhancement filtering is based on the poles of the LPC vocal tract filter to improves the match between synthetic and natural bandpass waveforms for a more natural quality to the speech output.
- MELPe supports multiple rates of 600 bps, 1200 bps, and 2400 bps with dynamic transcoding between rates.
- MELPe Vocoder mixed-excitation model represents a richer range of speech characteristics and produces more natural sounding speech.
- MELPe algorithm is very efficient for use in low power applications.
- MELPe is available as binary or source code (C and assembly) either standalone, as part of a library, or with a VoIP stack.
- MELPe may be licensed for a flat fee and VOCAL does not charge a royalty