Humans are very good at recognizing speech: we can understand all speakers of our language, even people we"ve never heard before. Computers are very bad at it: at best, can recognize very general speech, or the speech of one person, even though computers are pretty good at other tasks. These waves have several important properties for example, pitch and volume. Changing the amplitude of a sound wave changes its volume. We can analyze the acoustic properties of signals by looking at a spectrogram: shows the amplitude of a waveform at different frequencies. Speech is made up of areas of concentrated acoustic energy called formants. Formants: bands of high-amplitude frequencies in speech formant transitions: consonants: steady-state formants: vowels. We know a lot about the acoustics of speech. But the task of decoding speech isn"t as simple as it seems lack of invariance: speaker variability, segmentation. Scientists tried cutting a recording of speech into separate phonemes.