초록 ▼

An acoustic detection and localization system includes a signal acquisition stage, a biomimetic processor, and an acoustic feature processor. The system uses multiple acoustic cues including spectral content, inter-aural time delay (ITD), inter-aural intensity difference (IID), and periodicity cont

An acoustic detection and localization system includes a signal acquisition stage, a biomimetic processor, and an acoustic feature processor. The system uses multiple acoustic cues including spectral content, inter-aural time delay (ITD), inter-aural intensity difference (IID), and periodicity content to detect, classify, and localize sound sources. Pairs of acoustic sensors are arranged geometrically with a spacing depending on the application. The biomimetic processing provides for echo suppression to enhance performance in reverberant environments, and automatic gain controls for robustness in noisy environments. Applications include mobile robotic platforms for reconnaissance and surveillance, helmet mounted systems for sniper detection, vehicle-mounted systems for combat awareness, general civilian security systems, and systems for environmental monitoring and tracking of animals.

대표청구항 ▼

What is claimed is: 1. A biomimetic processing system operative to process sensor signals in a biomimetic fashion to extract acoustic features therefrom, the sensor signals including a source acoustic signal of interest, comprising: (A) a biomimetic preprocessor, including for each sensor signal: (

What is claimed is: 1. A biomimetic processing system operative to process sensor signals in a biomimetic fashion to extract acoustic features therefrom, the sensor signals including a source acoustic signal of interest, comprising: (A) a biomimetic preprocessor, including for each sensor signal: (1) a filter bank operative to generate a plurality of frequency channels from the sensor signal; (2) for each frequency channel: (a) compression/rectification circuitry; (b) AGC circuitry; and (C) multi-threshold spike generation circuitry operative to generate a plurality of streams of spikes for a corresponding plurality of distinct sound level thresholds, each spike being a pulsatile electrical signal analogous to a biological action potential generated by a nerve fiber; (B) one or more feature extracting modules each employing event-based processing on the streams of spikes to extract corresponding acoustic features; and (C) a feature integration module operative in response to the acoustic features from the feature extracting modules to generate an output identifying a location of the source acoustic signal of interest. 2. A biomimetic acoustic detection system for use in an operating environment, comprising: an acoustic acquisition system including a plurality of acoustic sensors in an application-dependent geometric arrangement and being operative to generate respective sensor signals representing respective acoustic inputs to the acoustic sensors; a biomimetic processing system according to claim 1; and a feature processing sub-system operative to classify the acoustic features extracted by the biomimetic processing sub-system to identify predetermined acoustic sources and their respective locations in the operating environment. 3. A biomimetic processing system according to claim 1, wherein: the feature extracting modules include a plurality of inter-sensor time difference (ITD) modules operative to compare spikes of the same frequency bands and thresholds across each pair of the sensors, each ITD module including (1) a coincidence detector acting as a counter, the counter being start-triggered by a spike of one of the sensors of a pair and stop-triggered by a spike of the other sensor of the pair, and (2) a timeout circuit operative to generate an ITD event signal if a time counted by the counter falls within a predetermined time window corresponding to a geometry of the sensors, and to otherwise refrain from generating the ITD event signal; and the feature integration module is operative to combine the respective ITD event signals of the ITD modules to derive an indication of direction of the source acoustic signal of interest, the ITD event signals of the respective sound level thresholds being combined in a first weighted fashion to realize a desired dynamic range, the ITD event signals of the respective frequency channels being combined in a second weighted manner to realize a desired signal to noise ratio. 4. A biomimetic processing system according to claim 3, wherein: the number of pairs of sensors is greater than one, and the ITD event signals of each pair of sensors indicate a respective cone of confusion for the direction of the source acoustic signal; and the ITD event signals of the respective pairs of sensors are combined to find intersections of the cones of confusion to remove spatial ambiguities in the direction of the source acoustic signal. 5. A biomimetic processing system according to claim 1, wherein the feature extracting modules include a plurality of inter-sensor intensity difference (IID) modules operative to compare spikes of the same frequency bands across each pair of the sensors, each IID module including a pair of adder circuits for the spikes of a respective pair of sensors, each adder circuit generating a respective weighted sum of the spikes for the respective sound level thresholds, each IID module further including a differencing circuit operative to compare the weighted sums from the adder circuits. 6. A biomimetic processing system according to claim 1, wherein the feature extracting modules include a plurality of periodicity modules operative to compare time-separated spikes of the same frequency bands, thresholds and sensors, each periodicity module including a coincidence detector acting as a counter, the counter being start-triggered by a first spike of a sensor and stop-triggered by a temporally adjacent second spike of the same sensor. 7. A biomimetic processing system according to claim 1, wherein the AGC circuitry comprises a cascade of AGCs including a first AGC and second AGC, the first AGC having a relatively fast adaptation time and the second AGC having a relatively slow adaptation time, the combination of fast and slow adaptation times being effective to emphasize fast transients and signal onsets while suppressing multiple transients and longer term signal variations that may be associated with reverberation. 8. A biomimetic processing system according to claim 1, wherein the multi-threshold spike generation circuitry includes a low-threshold spike generating circuit, a medium-threshold spike generating circuit, and a high-threshold spike generating circuit, each spike generating circuit including a respective threshold crossing detector, an absolute refractory circuit, and a relative refractory circuit, each threshold crossing detector being operative to compare a signal from the AGC circuitry to a respective threshold value to generate a corresponding spike output. 9. A biomimetic processing system according to claim 8, wherein the threshold value of the low-threshold spike generating circuitry is approximately 20% above a quiescent output value of the AGC circuitry, the threshold value of the medium-threshold spike generating circuitry is approximately 10% above a maximum saturated steady-state output value of the AGC circuitry, and the threshold value of the high-threshold spike generating circuitry is approximately 40% of a maximum peak output value of the AGC circuitry.