# US 2025/0095643 A1 · Low Power Always-on listening Artificial Intelligence (AI) System > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 2025/0095643 A1 - Document ID: US20250095643A1 - Status: Published application - Kind: A1 - Application number: 18/468,964 - Filed: 2023-09-18 - Published: 2025-03-20 - Earliest claimed: 2023-09-18 - Assignees: QUALCOMM Incorporated - Applicants: QUALCOMM Incorporated - Inventors: Kyu Woong HWANG; Simyung CHANG; Sungha CHOI; Joo Seong JEONG; Hyoungwoo PARK; Kyuhong SHIM - Web page: https://patentgrep.com/patent/US20250095643A1 ## CPC - G10L 15/197: Probabilistic grammars, e.g. word n-grams - G10L 15/063: {Training} - G10L 15/22: Procedures used during a speech recognition process, e.g. man-machine dialogue - G10L 2015/0635: {updating or merging of old and new templates; Mean values; Weighting} - G10L: SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING ## Abstract Various embodiments include systems and methods for continuous speech monitoring artificial intelligence solutions. A low-power always-on listening module (LPALM) may maintain continuous auditory awareness or alertness without consuming an excessive amount of the processing, memory, or battery resources of the user computing system or device. As such, the LPALM may operate on the computing device for an extended period of time without depleting the device's battery resources, rendering the user device non-responsive, or otherwise having a negative or user-perceivable impact on the performance, functionality, or power consumption characteristics of the user device. ## Claims ### Claim 1 A computing device, comprising: a processing system configured to: collect, in a low-power always-on listening mode, ambient audio data and storing the collected ambient audio data and activation timestamps in an audio buffer as buffered audio data; determine a confidence score based on a result of analyzing the buffered audio data for linguistic cues or non-linguistic cues; and activate a high-power listening mode and providing a last portion of the audio buffer to the high-power mode for instant context-aware assistance in response to determining that the confidence score exceeds a threshold value. ### Claim 2 (depends on 1) The computing device of claim 1, wherein the processing system is configured to determine the confidence score based on a result of analyzing the buffered audio data for linguistic cues or non-linguistic cues by determining the confidence score based on a result of analyzing the buffered audio data for linguistic cues that include syntax cues, semantic cues, subword cues, context cues, and co-occurrence cues. ### Claim 3 (depends on 1) The computing device of claim 1, wherein the processing system is configured to determine the confidence score based on a result of analyzing the buffered audio data for linguistic cues or non-linguistic cues comprises determining the confidence score based on a result of analyzing the buffered audio data for non-linguistic cues that include prosody cues, pitch cues, speech rate cues, volume cues, temporal pattern cues, and acoustic feature cues. ### Claim 4 (depends on 1) The computing device of claim 1, wherein the processing system is further configured to use a trained language model to identify linguistic cues or non-linguistic cues. ### Claim 5 (depends on 1) The computing device of claim 1, wherein the processing system is further configured to transition to a periodic training mode in response to determining that the computing device is connected to a stable power source. ### Claim 6 (depends on 1) The computing device of claim 1, wherein the processing system is further configured to: retrieve the buffered audio data and activation timestamps from the audio buffer; use a result of applying the retrieved audio data to a large language model (LLM) to identify instances in which activation high-power listening mode should have occurred but did not; label the identified instances and their corresponding timestamps; generate updated machine learning models for the low-power always-on listening mode based on the labels; and replace a current machine learning model of the low-power listening mode with the generated updated machine learning models. ### Claim 7 A method of continuously monitoring speech for preemptive or context-aware answers to user queries, comprising: collecting, in a low-power always-on listening mode operating on a processing system in a computing device, ambient audio data and storing the collected ambient audio data and activation timestamps in an audio buffer as buffered audio data; determining a confidence score based on a result of analyzing the buffered audio data for linguistic cues or non-linguistic cues; and activating a high-power listening mode and providing a last portion of the audio buffer to the high-power mode for instant context-aware assistance in response to determining that the confidence score exceeds a threshold value. ### Claim 8 (depends on 7) The method of claim 7, wherein determining the confidence score based on a result of analyzing the buffered audio data for linguistic cues or non-linguistic cues comprises determining the confidence score based on a result of analyzing the buffered audio data for linguistic cues that include syntax cues, semantic cues, subword cues, context cues, and co-occurrence cues. ### Claim 9 (depends on 7) The method of claim 7, wherein determining the confidence score based on a result of analyzing the buffered audio data for linguistic cues or non-linguistic cues comprises determining the confidence score based on a result of analyzing the buffered audio data for non-linguistic cues that include prosody cues, pitch cues, speech rate cues, volume cues, temporal pattern cues, and acoustic feature cues. ### Claim 10 (depends on 7) The method of claim 7, further comprising using a trained language model to identify linguistic cues or non-linguistic cues. ### Claim 11 (depends on 7) The method of claim 7, further comprising transitioning to a periodic training mode in response to determining that the computing device is connected to a stable power source. ### Claim 12 (depends on 7) The method of claim 7, further comprising: retrieving the buffered audio data and activation timestamps from the audio buffer; using a result of applying the retrieved audio data to a large language model (LLM) to identify instances in which activation high-power listening mode should have occurred but did not; labeling the identified instances and their corresponding timestamps; generating updated machine learning models for the low-power always-on listening mode based on the labels; and replacing a current machine learning model of the low-power listening mode with the generated updated machine learning models. ### Claim 13 A non-transitory computer readable storage medium having stored thereon processor-executable software instructions configured to cause a processing system in a computing device to perform operations for continuously monitoring speech for preemptive or context-aware answers to user queries, the operations comprising: collecting, in a low-power always-on listening mode operating, ambient audio data and storing the collected ambient audio data and activation timestamps in an audio buffer as buffered audio data; determining a confidence score based on a result of analyzing the buffered audio data for linguistic cues or non-linguistic cues; and activating a high-power listening mode and providing a last portion of the audio buffer to the high-power mode for instant context-aware assistance in response to determining that the confidence score exceeds a threshold value. ### Claim 14 (depends on 13) The non-transitory computer readable storage medium of claim 13, wherein the stored processor-executable software instructions are configured to cause a processor to perform operations such that determining the confidence score based on a result of analyzing the buffered audio data for linguistic cues or non-linguistic cues comprises determining the confidence score based on a result of analyzing the buffered audio data for linguistic cues that include syntax cues, semantic cues, subword cues, context cues, and co-occurrence cues. ### Claim 15 (depends on 13) The non-transitory computer readable storage medium of claim 13, wherein the stored processor-executable software instructions are configured to cause a processor to perform operations such that determining the confidence score based on a result of analyzing the buffered audio data for linguistic cues or non-linguistic cues comprises determining the confidence score based on a result of analyzing the buffered audio data for non-linguistic cues that include prosody cues, pitch cues, speech rate cues, volume cues, temporal pattern cues, and acoustic feature cues. ### Claim 16 (depends on 13) The non-transitory computer readable storage medium of claim 13, wherein the stored processor-executable software instructions are configured to cause a processor to perform operations further comprising using a trained language model to identify linguistic cues or non-linguistic cues. ### Claim 17 (depends on 13) The non-transitory computer readable storage medium of claim 13, wherein the stored processor-executable software instructions are configured to cause a processor to perform operations further comprising transitioning to a periodic training mode in response to determining that the computing device is connected to a stable power source. ### Claim 18 (depends on 13) The non-transitory computer readable storage medium of claim 13, wherein the stored processor-executable software instructions are configured to cause a processor to perform operations further comprising: retrieving the buffered audio data and activation timestamps from the audio buffer; using a result of applying the retrieved audio data to a large language model (LLM) to identify instances in which activation high-power listening mode should have occurred but did not; labeling the identified instances and their corresponding timestamps; generating updated machine learning models for the low-power always-on listening mode based on the labels; and replacing a current machine learning model of the low-power listening mode with the generated updated machine learning models. ### Claim 19 A computing device, comprising: means for collecting, in a low-power always-on listening mode, ambient audio data and storing the collected ambient audio data and activation timestamps in an audio buffer as buffered audio data; means for determining a confidence score based on a result of analyzing the buffered audio data for linguistic cues or non-linguistic cues; and means for activating a high-power listening mode and providing a last portion of the audio buffer to the high-power mode for instant context-aware assistance in response to determining that the confidence score exceeds a threshold value. ### Claim 20 (depends on 19) The computing device of claim 19, wherein means for determining the confidence score based on a result of analyzing the buffered audio data for linguistic cues or non-linguistic cues comprises means for determining the confidence score based on a result of analyzing the buffered audio data for linguistic cues that include syntax cues, semantic cues, subword cues, context cues, and co-occurrence cues. ### Claim 21 (depends on 19) The computing device of claim 19, wherein means for determining the confidence score based on a result of analyzing the buffered audio data for linguistic cues or non-linguistic cues comprises means for determining the confidence score based on a result of analyzing the buffered audio data for non-linguistic cues that include prosody cues, pitch cues, speech rate cues, volume cues, temporal pattern cues, and acoustic feature cues. ### Claim 22 (depends on 19) The computing device of claim 19, further comprising means for using a trained language model to identify linguistic cues or non-linguistic cues. ### Claim 23 (depends on 19) The computing device of claim 19, further comprising means for transitioning to a periodic training mode in response to determining that the computing device is connected to a stable power source. ### Claim 24 (depends on 19) The computing device of claim 19, further comprising: means for retrieving the buffered audio data and activation timestamps from the audio buffer; means for using a result of applying the retrieved audio data to a large language model (LLM) to identify instances in which activation high-power listening mode should have occurred but did not; means for labeling the identified instances and their corresponding timestamps; means for generating updated machine learning models for the low-power always-on listening mode based on the labels; and means for replacing a current machine learning model of the low-power listening mode with the generated updated machine learning models.