The routines provided here are:
melfcc.m - main function for calculating PLP and MFCCs from sound waveforms, supports many options - including Bark scaling (i.e. not just Mel! but cannot do rasta).
invmelfcc.m - main function for inverting back from cepstral coefficients to spectrograms and (noise-excited) waveforms, options exactly match melfcc (to invert that processing).
rastaplp.m - the original main routine to convert waveform data into a sequence of feature frames. Outputs are both cepstra and spectra features, and options allow for selection of RASTA, PLP, both, or neither. (Cannot do Mel-scaling). m = rastaplp(d,sr,0,ord) is equivalent to m = melfcc(d,sr,'preemph',0,'modelorder',ord,'numcep',ord+1,'dcttype',1,'dither',1,'nbands',ceil(hz2bark(sr/2))+1,'fbtype','bark','usecmp',1);
powspec.m - calculate the short-time power spectrum, basically a wrapper around Matlab's specgram.
audspec.m - map the power spectrum to an auditory frequency axis, by combining FFT bins into equally-spaced intervals on the Bark axis (or one approximation of it).
fft2barkmx.m - function to create the weight matrix that maps FFT bin magnitudes to the Bark frequency axis, used by audspec.m.
fft2melmx.m - generates a matrix of weights to convert FFT magnitudes into Mel bands, just like fft2barkmx above.
rastafilt.m - filter each frequency band (now in terms of log energy) with the RASTA filter.
postaud.m - fix-up the auditory spectrum with equal-loudness weighting and cube-root compression.
dolpc.m - convert the auditory spectra directly to LPC coefficients via Levinson-Durbin.
lpc2cep.m - convert LPC coefficients directly to cepstral values.
lpc2spec.m - convert LPC coefficients back into spectra by sampling the z-plane.
spec2cep.m - calculate cepstra by taking the DCT/DFT of the log of a set of spectra.
hz2bark.m - convert frequency in Hz to the auditory Bark scale.
bark2hz.m - convert back from Bark units to frequency in Hz.
hz2mel.m - convert frequency in Hz to the auditory Mel scale (either Slaney's or HTK mapping).
mel2hz.m - convert back from Mel units to frequency in Hz.
lifter.m - apply (or remove) weighting from cepstral dimensions.
deltas.m - calculate delta features over a limited window, just like feacalc/calc_deltas etc.
process_options.m - Mark Paskin's utility to parse long 'name', value pair lists (which I found out about through Kevin Murphy's KPMtools), used by melfcc.m.
cep2spec.m - inverse of spec2cep, undoes the DCT.
invpowspec.m - invert powspec.m i.e. go back from an STFT magnitude to a (noise-excited) time waveform.
ispecgram.m - precisely invert the short-time Fourier transform performed by specgram, taking the same argument (but fudges inverting the window at the moment).
invaudspec.m - invert audspec i.e. expand the condensed, nonlinear frequency axis to the full FFT detail. Intrinsically lossy, but does its best.
invpostaud.m - undo the weighting and compression of postaud, mostly lossless except the very edge bands are lost.
没有合适的资源?快使用搜索试试~ 我知道了~
资源推荐
资源详情
资源评论
收起资源包目录
音频信号处理 matlab代码.7z (27个子文件)
downed_code
fft2melmx.m 5KB
deltas.m 823B
postaud.m 2KB
lpc2cep.m 691B
lifter.m 955B
invmelfcc.m 2KB
fft2barkmx.m 1KB
bark2hz.m 536B
process_options.m 5KB
cep2spec.m 1KB
spec2cep.m 2KB
invpostaud.m 1KB
ispecgram.m 1KB
powspec.m 2KB
rastafilt.m 1016B
hz2bark.m 608B
rastaplp.m 2KB
help.txt 3KB
hz2mel.m 972B
audspec.m 1KB
invaudspec.m 1KB
dolpc.m 648B
lpc2spec.m 946B
readhtk.m 3KB
invpowspec.m 1KB
mel2hz.m 912B
melfcc.m 4KB
共 27 条
- 1
资源评论
- z4514656452018-01-19为什么老是没有下载成功,我的网速不好吗
- wangwenwei19862020-06-23首先感谢分享,还在看怎么使用
- m0_660854642022-04-12能不能加上详解啊,这根本看不懂
c1j2x3
- 粉丝: 2
- 资源: 4
上传资源 快速赚钱
- 我的内容管理 展开
- 我的资源 快来上传第一个资源
- 我的收益 登录查看自己的收益
- 我的积分 登录查看自己的积分
- 我的C币 登录后查看C币余额
- 我的收藏
- 我的下载
- 下载帮助
安全验证
文档复制为VIP权益,开通VIP直接复制
信息提交成功