This chapter presents a survey of standard and advanced methods for the analysis and modelling speech signals. First it introduces several speech processing functions as part of voice communication systems technology and proceeds to a brief description of human speech production. From this, a two-tier physical model of speech emerges which embraces the speech organ movements at the articulatory tier and the coupled aerodynamic flow and sound propagation at the aero-acoustic tier. Both of these physical tiers appear as separate components in most computational speech signal models. Their discussion addresses both the standard view of linear short-time stationarity and more advanced concepts from non-stationary processes (underspread processes, cyclostationarity) and non-linear systems (neural networks, non-linear oscillators).
展开▼
