Problem to plot raw pitch from a PC recorded audio file

javg · January 23, 2023, 3:49am

Hi, for the last two days I have been playing with the features in Speech Analyzer to become familiar with them. In particular, I’m trying to plot the raw pitch for some PC-recorded audio files, but I’m getting the following message every time I select the command to plot raw pitch:

“No voicing found in the range of human speech.”

I have encountered no problems to plot spectrogram of these PC-recorded .wav files, but the formants are not being plotted despite I have selected the option in the right-click Parameters… window.

I have been able to plot Intensity as well, but not the rest of the wide range of graph options available in Speech Analyzer.

I could be missing something but after reviewing what I have done and still being stuck at it, I decided to ask for help.

Before starting this new thread, I have conducted the following troubleshooting:

Downloaded and checked the Student Manual looking for a troubleshooting guide. (I found no reference to this problem)
Typed “No voicing found in the range of human speech.” in Google. (I found no reference to this problem).
Searched among the existing inquires in this online community. (I found no reference to this problem).
Tested the raw pitch plot function with three different PC-recorded audio files. (UNSUCCESSFUL).
Tested the raw pitch plot function with two different iPhone-recorded audio files. (UNSUCCESSFUL).
Tested the raw pitch plot function using three of the example audio files that come with the installation of Speech Analyzer. (SUCCESSFUL).

In summary, the option to plot raw pitch and formant (and few other graphs) does not work on any .wav file other than the examples .wav files that are part of Speech Analyzer installation.

Could you think of any reason why I can’t plot raw pitch and formants in the files I’m recording and opening with Speech Analyzer?

Thank you for any guidance you could provide.

Jorge

Matthew_Lee · January 23, 2023, 9:43pm

I’m interested in this question, as I’ve never had problems with SA with “normal” wave files. Not all files are created equal and SA expects a no-frills mono WAV file.
From the Help:

“Speech Analyzer only supports mono recordings and single channel (Type 0) MIDI files.”

I have a feeling that the files you are trying to open are not “standard” mono WAV files and are actually something stereo or compressed like an MP3 or AAC.

Could you share some of the wav files though Google or Dropbox?

What apps are you using to do the recordings? What export settings do you use? If not, could you give a screenshot of the descriptive properties of one of the wave files that won’t work.

Speech Analyzer:

Full Screen with Spectrograph:

Terry_Gibbs · January 26, 2023, 6:30am

Javg

Sorry about the trouble. Here are a few ideas…

#1
If you start SA and select your WAV file, you should see that waveform on the screen.
By default the 2 cursors at the far left of the recording now need to be positioned.
That is, move the Green and Red cursors to select a specific region of interest.
Then press one of the Triangle play buttons at the top of the screen.
You should now be able to hear the audio of the recording.

#2
In the left-hand “Phonetic” column, you should see several options to choose from.
If you select “Waveform, Auto Pitch” you will see a new screen under “Waveform”.
The Pitch contour of the audio recording should now be shown in that window.
A right-mouse click in the Auto Pitch graph will show the other 3 Pitch options:
Raw Pitch, Custom Pitch, and Smoothed Pitch.

#3
If you select Spectrogram, you can view the waveform, but no Formant will be shown.
To show Formants, do a right-mouse click in the Spectrogram and select Parameters.
Then select the item “Formant Tracks” and click the item “Apply” then select “Close”.

We have just released a new version of SA: “SpeechAnalyzer.3.1.2.0.exe”
I suggest you download it at: Speech Analyzer - SIL Language Technology - SIL Language Technology

Terry

Terry_Gibbs · January 26, 2023, 7:59am

Matthew_Lee

We’ve recently been doing updates in Speech Analyzer and it works well now.
SA supports recordings at 96KHz, 44KHz, and 22KHz either in Stereo or Mono.
However for linguistic work 22KHz Mono works well and results in a small file size.
There is little use for Stereo since our focus is getting clean audio with no crosstalk.

The comment of "Speech Analyzer only supports mono recordings and single channel
(Type 0) MIDI files.” is true for music analysis work. And SA only provides MP3 and WMA
support for read-only access, they would need to be converted before SA could use them.

Terry

javg · February 13, 2023, 5:39am

Hi Matthew,
It took me a while to get back to you, but here you have some comments:

The files I have recorded are .wav, MONO, sampled at 44100 Hz, 16 Bits. See attached screenshot. However, for some reason, SA does not report the file extension as shown in the examples you sent in your reply.
I recorded those files using a couple of iPhone Apps. The last one is Voice Recorder from TapMedia Ltd. See screenshot below.

I have tested multiple .wav files since I took back the issue yesterday. Now I see that the problem I reported is still present in some .wav files but other .wav files acquired with the same App are working just fine. I can’t find what the difference is (or are) between those files.

The file that continues to puzzle me is one called iPhone_600+1300hz.wav (2,684 KB):

I’ll work with the files that are not giving me trouble and will try to record the troublesome ones again.

Thank you for your help and interest.

Jorge

javg · February 13, 2023, 6:04am

Hi Terry,

Thank you for the ideas.
I prepared few pages with screenshots summarizing the additional troubleshooting I have completed in the last 36 hour. You will see that some files (2 out of 4 that I used to create the document) are giving trouble and SA can’t plot any Auto Pitch. While the other two files have no problem.

SA can’t plot Auto Pitch when I load a file that has 600 and 1300 Hz components. SA says:
“No voicing found in the range of human speech.”, but these frequencies are clearly in the range.

On the other hand, another .wav file with 580 Hz recorded with the same App and in the same way, has no issue to be analyzed by Auto Pitch.

I would have thought that if SA found “voicing” in a file with 580 Hz component, then it shoudl have found “voicing” as well in a file with 600 + 1300 Hz components. Please refer to attached file.

I thought that probably I was misunderstanding the meaning of “voicing”, so I looked it up:
Linguistics The vibration of the vocal cords during the production of speech or a speech sound."
Based on this definition, 600 and 1300 Hz should be recognized as vibrations produced by the vocal cords, as 580 Hz.

I’ll install the new version of SA and work with it from now on.
As I mentioned to Matthew, I will work with the files that are not giving me trouble.
I’ll leave the file iPhone_600+1300hz.wav (2,684 KB) aside.

Thanks again.

Jorge

javg · February 13, 2023, 6:05am

Slide 2 of 5

javg · February 13, 2023, 6:06am

Slide 3 of 5

javg · February 13, 2023, 6:06am

Slide 4 of 5

javg · February 13, 2023, 6:07am

Slide 5 of 5

chris_hirt · February 13, 2023, 12:35pm

I would suggest using ffprobe to analyze the different encoding, settings and structure of the two files to try and determine what the difference is.

ffprobe is a command-line tool, which you can download and use on your machine. If you don’t know how to use a command line tool, there are online services which will run ffprobe for you if you host the file somewhere. See https://ffprobe.a.video/

Terry_Gibbs · February 14, 2023, 4:30am

Jorge,

Thanks for writing in about this issue.

Your main comment was:
“SA can’t plot Auto Pitch when I load a file that has 600 and 1300 Hz components. SA says:
“No voicing found in the range of human speech.”, but these frequencies are clearly in the range.”

The reason that Speech Analyzer says “No voicing found in the range of human speech.” is because the default settings for Auto Pitch are intended for the narrow range of “Human Speech”. Speech Analyzer can do other things outside of just using Auto Pitch. Here are some comments regarding the use of Pitch.

By default SA shows the Auto Pitch graph. In the menu item Graph/ Parameter… it is common for a user to select either Linear or Logarithmic, and then select a region of interest bound by the users settings of Upper Boundary and Lower Boundary. Much of the work that is done with SA is focused on Phonetic analysis and thus analyzing sounds in the male voice range of 80 to 150HZ and female voice range of 160 to 250Hz is common.

However SA also has other Pitch settings that can be used for sound analysis which are in the 10Hz to 500Hz region and above. By selecting the main menu item Graphs/ Types then the 3 other Pitch options of Raw Pitch, Custom Pitch and Smoothed Pitch are shown.

Here is an example of Speech Analyzer using a higher frequency of 1300Hz.

There are also additional options that are listed in Tools/ Options. The options for “Cursor Alignment” are Glottal Pulse, Sample and Zero Crossing. The other options are the settings of the “Status Bar Units” of Time, Samples and Bytes.

Note:
In the main menu Help/ View Help there is this comment regarding the use of higher Pitch values.
This information is located in Search/ Pitch/ Pitch Tab, Graph Parameter.

"If the actual fundamental frequency is above the Max Frequency setting, the Pitch graphs might display a sub-harmonic frequency value. For example, if the actual frequency is 1200 Hz, the graphs might display a frequency of 600 Hz (1/2) or 400 Hz (1/3), depending the settings.

I hope this helps with the analysis work you’re doing.

Terry