[Simon]

You are telling HRest to load models/proto/$PROTO whereas it should start with the models just created by HInit, although this won’t actually cause an error in HTK.

Look for an error message from HRest – if it is failing to create any models, it should report an error.

Remember to always wipe all models (from hmm0 and hmm1) before every experiment: this will help you catch errors.

[Simon]

You should use either a loop (around all the files to be recognised), or the -S option (to pass the name of a file, in which all files to be recognised are listed), but not both.

The -S option will generally be faster(*). Why is that?

(*) although you might not notice the difference on the lab computers, because network speed typically dominates the run-time.

[Simon]

Yes, there are a total of 39 elements in the feature vectors. 12 of them are the MFCCs. You have 12+13+13=38 though.

The language model computes P(W) where W is the word sequence of one utterance to be recognised. For the digit recogniser, can you locate and inspect the language model that is being used?

The acoustic model computes P(O|W) where O is the observation sequence. How does O relate to the MFCCs?

How are P(O|W) and P(W) combined to calculate P(W|O), and why do we need to do that?

[Simon]

You should also exercise extreme caution in uploading data to external AI tools, since they are likely to retain this data and potentially include it in the training data for a future update of the tool.

You do not have permission to share any of the data for this assignment outside the University.

[Simon]

Please read “Please briefly describe if you use any external Artificial Intelligence (AI) related tools in doing this assignment…” which applies to all use of such tools.

Whilst professional programmers certainly use AI “co-pilots” to write code faster, I strongly recommend against this for beginners: for one thing, you do not yet have the skill to judge whether the resulting code is correct!

You will learn a lot more, and build more confidence, doing it yourself from scratch.

[Simon]

These notes are not currently available – there is everything you need here on speech.zone and the forums.

[Simon]

Reproducing a figure from another source may not be the best way to show your understanding.

If you really want to do this, then you need to cite the source in the caption of the figure, in the same way that you would cite the source of a direct text quote.

It’s up to you whether to include the original figure, or redraw it.

[Simon]

No, they are not. I’ve clarified that in the Formatting instructions.

[Simon]

You can check when AT 4.02 is available (i.e., whenever there is no class scheduled) by visiting timetables then searching for 4.02 and selecting the one in Appleton Tower. This link should take you directly there (although you may need to authenticate with EASE first).

[Simon]

Key points

You are correct that the x-axis (horizontal) will be frequency, and will be labelled in units of Hertz (Hz).

The vertical axis is magnitude, which is most commonly plotted on a logarithmic scale and is therefore labelled in decibels (dB).

Additional detail

Magnitude is a ratio (in this case, of filter output to its input), and therefore has no units: formally, we say it is dimensionless. So dB are not actually a unit, but a scale.

[Simon]

emulabel is an outdated program mentioned in some old documentation on festvox.org

In Pitchmark the speech, the command make_pmlab_pm creates label files from the pitchmarks, and places them in the pm_lab directory. These can be viewed in the same way as any other label file (such as the aligned phone labels), e.g., using wavesurfer.

[Simon]

You can use Qualtrics to build the survey, but host your audio files somewhere else, then enter their URLs into Qualtrics.

You can host the audio files anywhere that is able to provide a URL for the file. For example, a free github page, which might give you URLs like this:

https://jonojace.github.io/IS19-robustness-audio-samples/figure3/g_100clean.wav

[Simon]

Yes, you need both an abstract an an introduction.

[Simon]

You need to more clearly separate two independent design choices:

1. How to estimate F0 for recorded speech (which will become the database for a unit selection system, or the training data for a FastPitch model).

The method for estimating F0 (whether autocorrelation based like RAPT, or something else) is independent of the method used for synthesis. The synthesis methods just need values for F0, they don’t care where they come from.

2. Using F0 during synthesis (which will be either the unit selection algorithm, or FastPitch inference).

In a unit selection system that doesn’t employ any signal modification, you are correct in stating that the system can only synthesise speech with F0 values found in the database. FastPitch can, in theory, generate any F0 value.

But both methods use the data to learn how to predict F0, so they are both constrained by what is present in the database. The ‘model’ of F0 prediction in unit selection is implicit: the combination of target and join cost function. The model of F0 prediction in FastPitch is explicit.

So, in practice, as you suggest, FastPitch is very constrained by what is present in the training data. In that regard, it’s not so very different to unit selection.

[Simon]

There is probably either a formatting error or a non-ASCII character in your utts.data.

If you can’t easily locate it, try using binary search to find the offending line (here I’ll assume utts.data has 600 lines):

0. make a backup of utts.data

1. make a file containing only first half of utts.data, for example with

head -300 utts.data > first.data

2. try check_script on first.data

3a. if you get an error then take the first half again

head -150 utts.data > first.data

3b.if you don’t get an error, make a file containing the first three-quarters of utts.data

head -450 utts.data > first.data

and iterate, varying the number of lines in a binary search pattern, until you home in on the error.

[Author]

Posts