Spectrographic Analysis of Speech
Wideband spectrogram provides higher temporal resolution at the cost of spectral resolution.
Narrowband spectrogram provides higher spectral resolution at the cost of spectral resolution.
Wideband spectrogram uses a short window size (typically 5 ms) for analysis, which results in a broader main lobe of the spectrum of the window signal. This smears or smoothens the spectral features of the speech segment thereby reducing the spectral resolution.
Periodicity information of voiced sounds is reflected as vertical striations in WB spectrogram, while the pitch harmonics manifest as horizontal striations in NB spectrogram.
Spectrogram reading (i.e., to decipher the message from the spectrogram alone) is possible, but requires a lot of training.
There are transitions among sound units, which can be used for marking boundaries between the units.