introduce

Voicebox is a speech processing toolbox, including MATLAB routines, which is mostly written by Mike Brooks, Department of Electrical and Electronic Engineering, Imperial College London, sw7 2bt Exhibition Road, London, UK. Several routines need V5 of MATLAB.

This routine can be used as a compressed tar file or ZIP file, and the license in GNU is provided according to the terms of the contract.

Please send any comments, suggestions, error reports, etc. To mike.brookes @ ic.ac.uk

catalogue

Audio file input/output

Read and write WAV and other voice file formats

frequency table

Switch between Hertz, Mel, Employees Retraining Board and MIDI Frequency Table

Fourier/discrete cosine transform/Hartley transform

Various correlation transformations

Random number generation

Generate random vectors and noise signals.

Vector distance

Calculate the distance between vector lists.

Language analysis

Activity level estimation, expressed by spectrogram.

LPC analysis speech

Linear predictive coding program

Speech synthesis

Glottal waveform mode

Speech enhancement

Spectral noise subtraction

speech coding

Pulse code modulation, vector quantization.

speech recognition

Front-end processing approval

utility function

Miscellaneous practical functions

Audio file input/output

Routine can be read and, in some cases, written in various file formats:

Read and write suffix

Readwav writewav .WAV, etc. These routines allow any number of channels, and can also handle linear PCM (any accuracy up to 32 bits), conventional PCM and wood PCM. Large files can be read and written in small pieces.

Supplementary question: readhtk writehtk. Entropy hidden Markov toolbox used by HTK liquid to read and write waveform files.

Readsfs. Self-supporting voice archiving system files, by Mark Huckwall in UCL.

Readsph .NIST format file for mobile phone field (including timit).

Read AIF. AIF audio interchange file format used by MAC users.

Large-scale frequency reconstruction

Mel table is based on human perception of sine wave football field. The scale and frequency of this routine between mel2frq and frq2mel conversion is in Hz.

Employees Retraining Board, on a large scale, is based on equivalent rectangular bandwidth. Routine erb2frq and frq2erb conversion retraining board scale and frequency Hertz between interest rates.

Using MIDI standard, it is also clearly stipulated that C is plus 60 in semitones. This routine is used to convert between frq2midi and midi2frq, and between music frequency range and Hertz. In addition to the output, Frq2midi will also notice the character format in the name. Midi2frq can practice with the same scale, otherwise the Pythagorean scale is just a sound.