Identify words using time-frequency characteristics.

Good day, my task is to do word identification. I have used the pspectrum spectrogram function in order to aquire the power, time and frequency of a word. I was told to use the time-frfequency characteristics obtained and compare it to the time frequency characteristics of other words. This however, does not work for me as I constantly obtain incorrect results. The time frequency graph I obtain is often a linear graphs with gradients varieng by so little for every word. Example:

Does anyone know what characteristics of a sound wave can be used to compare to others' and to find out which word is the most similar/the correct word? I've tried so many things resulting in failure. Thanks!

Answer 1

[Pavan Guntha]

Hello Leon,

You could have a look at the following resources regarding interpreting basic time-frequency signal analysis.

1. Practical Introduction to Time-Frequency Analysis - Illustrates Time-Frequency analysis with a practical example of identifying numbers in a Dual-Tone Multi-Frequency (DTMF) signal.
2. Time-Frequency Gallery - It provides an overview of the time-frequency analysis features available in Signal Processing Toolbox and Wavelet Toolbox along with few examples.

Hope this helps!