Methods and apparatus for acoustic transient processing
AssigneeThe Johns Hopkins University
InventorsFernando J. Pineda, Gert Cauwenberghs, R. Timothy Edwards
Life of this patent
Estimating the term…
Abstract
Method and apparatus for using current-mode analog circuits to compute the correlation between an auditory input signal in the time-frequency domain and a stored binary template. By using massively parallel computation, circuits made of relatively imprecise analog components are able to accurately classify transient events, are competitive with high-performance DSP systems, and operate with much smaller power requirements.
Claims8
A mixed-mode very-large-scale integration (VLSI) device for real-time classification of acoustic transients by correlating an input signal comprising a plurality of input currents with a plurality of stored template values, the device comprising:
- means for generating the plurality of input currents from collected acoustic transient data;
- a normalizer for normalizing the plurality of input currents;
- a switched-capacitor circuit for time and/or channel differencing the input currents;
- a correlator array comprising a plurality of cells for receiving the plurality of normalized, time and/or channel differenced input currents and for comparing the input current received in each cell with the template value stored in each cell and for adding the input current to a total current if there is a match;
- a bucket brigade device for summing the input currents added by the cells of the correlator array;
- and
- means for classifying the acoustic transients using an output of the bucket brigade device.
A method for real-time classification of acoustic transients in mixed-mode very-large-scale integration (VLSI) comprising the steps of:
- collecting acoustic transient data;
- generating the input signal using the acoustic transient data;
- normalizing the input signal;
- computing the time and/or channel difference of the input signal;
- taking a time difference of a plurality of stored template values, a pairwise difference between neighboring channels, or a combination thereof;
- binarizing the stored template values;
- correlating the binarized, stored template values with the normalized time and/or channel differenced input signal;
- and
- classifying the acoustic transients using an output of the correlating step.
Dependencies are inferred from the claim text.