[Simon]

You have two separate problems there.

The first is probably just the keyboard mapping for the VM. Try typing @ on your keyboard and you will probably get ” (for me this is shift plus the 2 key).

The other sounds like a common audio problem and the solution is here.

[Simon]

You are right: most research papers have rather vacuous introductions such as “The popularity of speech synthesis has been rising in recent years.” or other waffle. Do not do that in coursework – there are no marks for this!

So, how to do better? Don’t simply omit the Introduction. Use it to say something meaningful about what is coming up in your report. Have you got a particular approach to the assignment? Did you find an original angle? What are your concrete goals (and don’t just restate the instructions, but give your interpretation). Are there questions you want answers to?

Use an introduction to set up some expectations in the mind of the reader, then make absolutely sure to deliver them later in the paper. Pose questions in the Introduction but don’t give answers. Motivate the reader to want to read on, to find out the answer. Use the Conclusion to remind the reader what was promised and how you have delivered it, leaving them feeling satisfied.

See also this post.

[Simon]

Getting closer… some more clues:

The lowest value of k (apart from 0), is 1. That’s the lowest frequency DFT “bin” and corresponds to the lowest frequency basis function.

The lowest frequency basis function makes just one cycle within the analysis frame. So its frequency in Hertz must be 1 / “duration of the analysis frame in seconds”

We can calculate the duration of the analysis window in seconds if we know how many samples are in the frame and what the sampling frequency is: there is a simple relationship between these three values.

Now we know the frequency of the first bin k=1 we can work out the frequency of the bin k=2 and so on all the way up to K.

[Simon]

In the magnitude spectrum (i.e., the frequency domain), we see the frequencies present in whatever signal we are analysing: the analysis frame. In the video and that time point, the analysis frame is the visible portion of waveform on the left, which has a duration of 30 ms and contains multiple pitch periods.

You can use the DFT to analyse any amount of waveform (= the analysis frame) that you wish. If you analyse only one pitch period, you will not see any evidence of periodicity. If you analyse a waveform that contains multiple pitch periods (like in the video at the time point) then you will see evidence of periodicity in the magnitude spectrum: harmonics.

[Simon]

You have two separate problems. The solution to the choppy sound issue is probably this.

The “Feature Word not defined” error means that either

a) you are not using the correct voice (which is loaded by config.scm)

or

b) you omitted a step in the pipeline

Don’t change the PATH back to what it was before – this will pick up the old version of Festival, which will not work for the assignment.

[Simon]

Try also looking in the VMware settings for this virtual machine – there are some options to fit the VM to the window and so on – experiment with those.

The other possibility is that your host computer has a very high resolution screen – see if you can adjust its settings to something lower resolution, although that will affect the host operating system too. This does the trick for me on my Mac.

[Simon]

Yes, nearly. In TD-PSOLA, our representation of the filter is the impulse response. This is a single pitch period. A diphone is a sequence of pitch periods.

Why can’t we store a single pitch period to represent the filter for a complete diphone?

[Simon]

total_entropy = ( sum(left_counts)*entropy(left_counts) + sum(right_counts)*entropy(right_counts) ) / (sum(left_counts) + sum(right_counts))

[Simon]

Rename (or delete) the notebook and ‘git pull’ to get a fresh copy of the original.

[Simon]

Which cell of which notebook are you referring to?

[Simon]

At the end of the part you want to select, click and hold the left mouse button. Then move the mouse up – you’ll start selecting text and eventually the window will scroll upwards until you’ve selected everything you need.

(There should be a keyboard shortcut, but I can’t get it to work.)

[Simon]

Make sure you activate the slp python virtual environment before installing these packages, and then make sure you run jupyter notebook with that environment activated.

You can tell that an environment is active because the shell prompt changes to include “(slp)”.

[Simon]

The source-filter model is a conceptual model to help us understand speech signals, by creating model of how they are produced.

We can use this model in many different ways, to do many different things with speech signals. Sometimes we actually implement the model directly, other times we use it as motivation and use it indirectly.

A direct application would be to use the model itself to process speech. We would fit the filter to natural speech, thus finding suitable values for the coefficients of the difference equation defining the filter. Then, we might generate synthetic speech at an F0 and duration of our choice by constructing an appropriate impulse train as the excitation signal to input to the filter.

In this direct application, we manipulate source features (F0, duration) by changing the excitation signal. We can also independently manipulate the filter’s resonances (formant frequencies) by adjusting the coefficients of the difference equation. We have total control over the signal.

But we can also use the model indirectly – that is, without ever writing down the difference equation or inputting an excitation signal into an actual filter. TD-PSOLA is an example of this way of using the source-filter model. Instead of using the difference equation to define the filter, we instead represent the filter as its impulse response (which is a waveform). TD-PSOLA only offers partial control over the signal. We cannot use it to manipulate the filter’s resonances because the impulse response is a waveform, and not a set of coefficients in an equation.

[Simon]

In the same directory you started Festival from.

[Simon]

I’ve not seen this error, but it does sound related to disk space. If you hit serious problems, or think the VM has been corrupted, you can delete the VM entirely (and empty your Trash) and download a fresh copy.

[Author]

Posts