Process for digitizing speech
First Claim
1. A process for encoding speech comprising the steps of;
- (a) discovering the random zero crossing events of speech in real time by using a non-clipping means to detect zero crossings;
(b) using an output signal of said means to detect zero crossings to drive a means to generate uniform pulses which are narrow, of constant width and constant amplitude, said pulses being time coincident with said zero crossings;
(c) transmitting said pulses to a remote point;
(d) receiving said pulses; and
(e) applying said pulses to an intelligent reconstruction means which is configured to latch up on receipt of a positive pulse and latch down on receipt of a negative pulse.
0 Assignments
0 Petitions
Accused Products
Abstract
A process is disclosed for digitizing (more precisely encoding) speech for the purpose of reducing information rate and bandwidth relative to that of prior art means to digitize speech, while enjoying a high signal to noise ratio. Using the the same encoding techniques the process can be used for storage of speech and machine recognition of speech. The process depends on detecting audio waveform zero crossings and generating uniform pulses at the time of the zero crossings. The uniform pulses are created through a regenerative process and are independent of the actual waveform shape save the time of zero crossing. Transmission (or storage) of these uniform pulses permits reconstruction of highly intelligible speech. Ratios of times between zero crossings is used in a new technique for machine word recognition; said ratios allowing recognition regardless of the speakers actual speech rate.
54 Citations
12 Claims
-
1. A process for encoding speech comprising the steps of;
-
(a) discovering the random zero crossing events of speech in real time by using a non-clipping means to detect zero crossings; (b) using an output signal of said means to detect zero crossings to drive a means to generate uniform pulses which are narrow, of constant width and constant amplitude, said pulses being time coincident with said zero crossings; (c) transmitting said pulses to a remote point; (d) receiving said pulses; and (e) applying said pulses to an intelligent reconstruction means which is configured to latch up on receipt of a positive pulse and latch down on receipt of a negative pulse.
-
-
2. A process for digitizing speech in preparation for transmission comprising the steps of;
-
(a) detecting the zero crossings of speech using a non-clipping means to detect zero crossings in real time; (b) using an output of said zero crossing means to drive a pulse generator the output of which are narrow constant width constant amplitude pulses; (c) applying said pulses to a time-to-analog converter to obtain an analog signal representing the time between said pulses; (d) applying said analog signals to an analog-to-digital converter to derive a series of parallel bit digital words representing the time intervals between zero crossings;
e) presenting said parallel bit digital words to a number of shift registers, each of said shift registers corresponding, respectively, to one bit position in the parallel word,f) keeping track of which cell of said shift registers has a bit stored in the most advanced position thereof by providing an auxiliary shift register, containing an inital one in which said "one" is advanced each time said analog-to digital converter completes a conversion, and provides a signal to said auxiliary shift register; g) advancing bits in said shift registers each time a conversion is completed by said analog-to-digital converter; h) connecting data lines from each of said parallel bit positions in said shift registers to commutators switched by said signals from auxiliary shift register; i) stepping the armatures of said commutators by means of a clock sequentially to sample bits from said data lines corresponding to each parallel bit position, said clock also driving a divide-by-four circuit; j) said divide-by-four circuit causing a backward shift of a one in said auxiliary shift register, closing the commutator switch position, and opening all other commutator switch positions and clearing any one in any shift register cell forward of said switch closure; k) providing a synchronizing signal from said clock to ensure reassembly of the now sequential bits in proper parallel order; and l) transmitting said now sequential bits at a constant average rate along with said synchronizing signal to a remote point. - View Dependent Claims (3, 4, 5)
-
-
6. A process for digitizing speech in preparation for storing in memory comprising the steps of;
-
a) detecting the zero crossings of speech waves to derive transition signals therefrom; b) using means for deriving data from said transition signals representative of said zero crossings to obtain, in real-time, time-interval pulses indicating the time intervals between successive zero-crossing events, said pulses being narrow, of constant amplitude and constant width; c) presenting said pulses to a time-to-analog converter; d) presenting the analog output of said converter to an analog-to-digital converter to produce digital words representing the time between zero crossings; e) providing a synchronization signal for the purpose of later separating said words, and f) storing said words in an appropriate memory. - View Dependent Claims (7)
-
-
8. A process for digitizing speech from several simultanious audio channels in preparation for transmission, comprising the steps of;
-
(a) detecting the zero crossing events in all of several audio channels simultaniously; (b) assigning a digital number to each of said channels; (c) transmitting to a remote point the digital number of a channel at the time said channel experiences a zero crossing event; (d) presenting said digital number to a decoding logic circuit comprising inverters and AND circuits, said AND circuits including outputs for each of said channels for which a digital number has been assigned; (e) said inverters and AND circuits being so connected that only one AND output will go up for a particular digital number presented, and only for that number corresponding to the channel said AND circuit is assigned to; (f) presenting said AND outputs to ballanced differentiators, the output of said differentiators being unidirectional, narrow constant width pulses; and (g) said unidirectional pulses being presented to a means for constructing speech which produces square waves, the periods of which correspond to the periods of said speech waves.
-
-
9. A process to encode speech suitably to form patterns for machine recognition of speech comprising the steps of;
-
a) discovering the random zero crossing event of speech in real time by using a non-clipping means to detect zero crossings; b) using an output signal of the means to detect zero crossings to drive a means to generate uniform narrow, constant width pulses, said pulses being time coincident with said zero crossings; c) presenting said pulses to a time-to-analog converter; d) using a first conversion period as a base period to be divided into successive converted times resulting in divide-by-base time analog values; e) converting the analog values to digital numbers; f) collecting a set of such numbers representing time ratios as time ratio patterns; g) preparing several such patterns to be stored in memory for a machine vocabulary.
-
-
10. A process for machine recognition of speech comprising the steps of;
-
a) detecting the zero crossing events in speech waves; b) gathering data defining the real-time between zero crossings by using a means for providing information on the time between zero crossings, said means producing narrow constant width pulses marking the real-time of said zero crossing events; c) presenting said pulses to a time-to-analog converter; d) preparing temporary time ratio patterns in a manner of the process used in claim 31 to create time ratio patterns; e) using said temporary patterns for comparison in a comparison means bit-by-bit with previously prepared time ratio patterns stored in appropriate memory as machine vocabulary; and f) upon a successful comparison having said comparison means provide a command to a device to carry out any of the functions assigned to a recognized word.
-
-
11. A process for pulse encoding speech comprising the steps of;
-
a) discovering the random zero crossing events of speech in real time by using a non-clipping means to detect zero crossings; b) using an output signal of said means to detect zero crossings to generate constant amplitude pulses the leading edge(s) of which are coincident with said zero crossing, the time length of said pulses may vary for the purpose of transmitting additional information and said pulses have a maximum allowable length;
said lengths having no bearing or relationship to the times between zero crossings; andc) transmitting said pulses to a remote point. - View Dependent Claims (12)
-
Specification