| 摘要 |
1321313 Helium speech processor NATIONAL RESEARCH DEVELOPMENT CORP 14 June 1971 [17 June 1970] 29400/70 Heading H4R In a speech processing system, e.g. for rendering helium speech intelligible, segments of speech signal are loaded into a storage device and subsequently removed therefrom at a different rate, the loading and unloading processes operating pitch synchronously during voiced sounds, and almost continuously during unvoiced sounds, so that the larynx periodicity of the original speech is preserved while shifting the significant vocal cavity resonant frequencies. As described the incoming speech after filtering and shaping to restore the spectral energy content level to that occurring in normal speech, from that occurring in speech in a high pressure environment, is converted to digital form, using a full wave rectifier 3 and polarity sampler 7 to provide digitization to 8 bit pcm relative the mean of the speech signal in the A to D converter 5. A peak picker 9 operates on the full wave rectified speech signal to provide a pulse 2À5 Msecs long to gate samples starting at the commencement of each pitch period into the four storage banks 10, the stores being filled in turn with successive pitch period samples of 2À5 Msec length, via the switch 6 which is stepped by the pitch period pulses from 9. The read out of the signals from the stores 10 is at the rate dictated by the pulse generator 16, and at a rate slower than the read in. The commutator switch 11 is stepped at four times the read shift rate so that a stored sample is read out from each bank of the store in turn before the read shift is actuated, and read out from the store currently being filled is inhibited. Normally only one store bank will have any data in it, the others having been emptied on a previous read cycle, but the arrangement allows overlap of two signal sections if the incoming pitch frequency and the length of each pitch period sample are such as to make the beginning of one sample overlap the end of a previous sample on the slow read out. The 8 bit pcm signals from the stores can be converted to analogue form and added, where more than one store has contents read out at the same time, or the adding can be digital with d to a conversion subsequently. The resulting signal is expanded, if compression has been applied before storage, and filtered to provide a signal which is more understandable than the original speech. |
