kpb: add multi-KPB WOV arbiter for 3-keyword DMIC capture with VAD by lgirdwood · Pull Request #11022 · thesofproject/sof

lgirdwood · 2026-07-22T12:06:35Z

EDIT now added test detection in each WoV module to wake on voice frequencies for male, female and children voice frequencies. i.e. each WoV will trigger on a different speaker type.

Adds a new WOV arbiter component and multi-KPB topology to support three simultaneous wake-word detectors sharing a single DMIC, with exclusive host PCM drain arbitration when any keyword fires.

New component: src/audio/wov_arbiter/

3 input pins (one per KPB host-sink), 1 output pin (to host copier)
AMS subscriber for AMS_WOV_DETECT_MSG_UUID: activates the triggering slot and broadcasts WOV_CTRL PAUSE to sibling detectors
On trigger(STOP): deactivates and broadcasts WOV_CTRL RESUME so all detectors return to listening
Idle inputs are silently drained to prevent buffer stalls
UUID 4a5b6c7d-8e9f-4a1b-2c3d-4e5f60718293 registered in uuid-registry

AMS additions (ams_msg.h):

AMS_WOV_DETECT_MSG_UUID: detector -> arbiter on keyword detect
AMS_WOV_CTRL_MSG_UUID: arbiter -> detectors for pause/resume control

detect_test extensions:

IPC4_DETECT_TEST_SET_WOV_SLOT large-config param assigns a slot id (0-2)
On detection: also sends AMS_WOV_DETECT_MSG_UUID with slot_id to arbiter
AMS consumer for WOV_CTRL: sets paused flag; copy() drains buffer without running the detector when paused
The existing energy-threshold detector fires on a hand clap, making detect_test a ready-to-use lab test stub for this topology

Topology: tools/topology/topology2/platform/intel/dmic-wov-multi.conf

Pipeline 100 (Core 0): DMIC Copier -> Mixin (fan-out to 3 Mixouts)
Pipelines 101-102 (Core 0): Mixout -> KPB -> detect_test (slots 0-1)
Pipeline 103 (Core 1): Mixout -> KPB -> detect_test (slot 2)
Pipeline 104 (Core 0): WOV Arbiter -> Host Copier (single PCM stream)
KPB output_pin_binding: pin 0 -> detector (sel_sink), pin 1 -> arbiter (host_sink)

Requires: CONFIG_COMP_WOV_ARBITER, CONFIG_COMP_KPB, CONFIG_COMP_MIXIN_MIXOUT,
CONFIG_AMS, CONFIG_COMP_KWD_DETECT

lgirdwood

Some opens.

lgirdwood

Needs more inline comments around code blocks and important statements can also do with more context in each commit message. I would put the README.md patch first to give reviewer context first before further review.

lgirdwood · 2026-07-29T15:27:56Z

 	}
 #endif /* CONFIG_IPC_MAJOR_4 */

+	if (!kpb->sel_sink || !kpb->host_sink) {


lets comment what this block is doing and why.

lgirdwood · 2026-07-29T15:28:08Z

 comp_err(dev, "could not find sink: sel_sink %p",
 kpb->sel_sink);
 ret = -EIO;
+	} else {


ditto re comments.

lgirdwood · 2026-07-29T15:29:01Z

-	#UUID: D8218443-5FF3-4A4C-B388-6CFE07B9562E
-	uuid "43:84:21:d8:f3:5f:4c:4a:b3:88:6c:fe:07:b9:56:2e"
+	#UUID: a8a0cb32-4a77-4db1-85c753d7ee07bce6 (kpb4 static module)
+	uuid "32:cb:a0:a8:77:4a:b1:4d:85:c7:53:d7:ee:07:bc:e6"


why do we need a new UUID

lgirdwood · 2026-07-29T15:30:04Z

 if (frames_to_copy + start_frame > mixout_data->mixed_frames)
 mixout_data->mixed_frames = frames_to_copy + start_frame;
+
+ if (mixout_mod && mixout_mod->dev)


need comment for what we are checking and why.

lgirdwood · 2026-07-29T16:09:59Z

@@ -0,0 +1,455 @@
+// SPDX-License-Identifier: BSD-3-Clause
+//
+// Copyright(c) 2024 Intel Corporation. All rights reserved.


lgirdwood · 2026-07-29T16:10:40Z

+ */
+	uint8_t active_slot;
+
+#if CONFIG_AMS


we should actually depend on AMS in teh Kconfig and then we can remove all these ifdefs

Not related to this PR, but I'm wondering whether we should/need to depend on AMS. It is currently disabled in many Intel configs.

lgirdwood · 2026-07-29T16:11:30Z

+	if(CONFIG_COMP_VAD_GATE)
+ add_subdirectory(vad_gate)
+	endif()


this should be in the vad gate patch.

lgirdwood · 2026-07-29T16:11:55Z

 e93326d8-0d14-4bf0-bcb9e063d3d80136 twb_sched
 42f8060c-832f-4dbf-b24751e961997b34 up_down_mixer
 6f6b6f4b-6f73-7466-20e1e62b9779f003 userspace_proxy
+8c7d6e5f-4a3b-2c1d-0e9f8a7b6c5d4e3f vad_gate


this should be in vad gate patch

lgirdwood · 2026-07-29T16:13:52Z

+ dai_type "DMIC"
+ copier_type "DMIC"
+ stream_name $DMIC1_NAME


isnt this passed in via build for the target ?

lyakh · 2026-07-30T06:49:10Z


 if (sink_buf_id == buf_id) {
+ struct comp_dev *sc = comp_buffer_get_sink_component(sink);
+ comp_err(dev, "kpb_bind: buf_id=%d sink_comp=0x%x -> %s",


is this an error condition?

lyakh · 2026-07-30T06:49:57Z

+	comp_dev_for_each_consumer(dev, sink) {
+ struct comp_dev *sc = comp_buffer_get_sink_component(sink);
+ if (sc) {
+ comp_err(dev, "kpb consumer in bsink_list: comp_id=0x%x type=%d sink_buf=%p",


replace comp_err() everywhere in this commit / PR

lyakh · 2026-07-30T06:50:53Z

+ if (sc) {
+ comp_err(dev, "kpb consumer in bsink_list: comp_id=0x%x type=%d sink_buf=%p",
+ dev_comp_id(sc), sc->drv ? sc->drv->type : -1, sink);
+ if (dev_comp_id(sc) != 0x10)


what is 0x10?

lyakh · 2026-07-30T06:51:58Z

+ kpb->sel_sink = sink;
+ else
+ kpb->host_sink = sink;
+ }


I suppose the loop should run exactly twice to only set each of these pointers once? Should we check or at least add a comment?

lyakh · 2026-07-30T06:52:39Z

+ if (!kpb->sel_sink)
+ kpb->sel_sink = sink;
+ else if (!kpb->host_sink)
+ kpb->host_sink = sink;


same here - shouldn't be overwriting, right?

lyakh · 2026-07-30T07:33:00Z

+}

-LOG_MODULE_REGISTER(kd_test, CONFIG_SOF_LOG_LEVEL);
+LOG_MODULE_REGISTER(kd_test, LOG_LEVEL_INF);


switch to CONFIG_SOF_LOG_LEVEL for consistency ?

lyakh · 2026-07-30T07:34:00Z

 	struct comp_data *cd = comp_get_drvdata(dev);

-	comp_info(dev, "entry");
+	comp_err(dev, "notify_host: WOV module_id=0x%x instance_id=0x%x slot_id=%u detected",


comp_err() wrong

lyakh · 2026-07-30T07:34:41Z

+	uint8_t slot_id = cd->wov_slot_id;
+	if (slot_id == WOV_SLOT_INVALID) {
+ if (dev->pipeline && dev->pipeline->pipeline_id >= 101 && dev->pipeline->pipeline_id <= 103)
+ slot_id = dev->pipeline->pipeline_id - 101;


what are these magic numbers?

lyakh · 2026-07-30T07:37:01Z

+ freq_match = (freq_hz >= 80 && freq_hz <= 170);
+ voice_type = "MALE";
+ break;
+ case 1: /* Female Voice Range: 175 - 270 Hz */


so 172Hz will be "nobody?" Or are frequencies detected with 5Hz granularity?

typical human frequency bands from internet, so yes - nobody.

@lgirdwood sorry, I cannot imagine this being right. Typical - yes. But I cannot imagine no man / woman / child being able to intersect frequencies. So, if someone can speak at 170Hz and somebody else can speak at 175Hz, I'm pretty sure it's humanly possible to speak at 172Hz too.

Its taken from some internet references, I'm not an expert on the subject so will leave it as is. Its not that important for the test.

lyakh · 2026-07-30T07:38:17Z

+	else if (ppl_id == 102 || ppl_id == 3)
+ cd->wov_slot_id = 1;
+	else if (ppl_id == 103 || ppl_id == 4)
+ cd->wov_slot_id = 2;


these numbers look familiar

Copilot

Pull request overview

This PR introduces a multi-slot Wake-on-Voice (WoV) capture design that allows three concurrent detectors to share a single capture source, with a new firmware “arbiter” component selecting which KPB drain feeds a single host PCM stream. It also adds a VAD gate component intended to suppress downstream processing during silence, and extends the detect_test sample to coordinate with the arbiter via AMS.

Changes:

Add new firmware components: wov_arbiter (multi-input drain routing + AMS control) and vad_gate (energy-based gating).
Add/adjust Topology2 definitions for multi-slot WoV, plus wiring/includes updates.
Extend detect_test for multi-slot behavior and AMS messaging; adjust rimage/module TOMLs for additional instances.

Reviewed changes

Copilot reviewed 38 out of 38 changed files in this pull request and generated 12 comments.

Show a summary per file

File	Description
uuid-registry.txt	Registers new UUIDs for `vad_gate` and `wov_arbiter`.
tools/topology/topology2/sof-hda-generic.conf	Adds includes for WoV/VAD/KPB-related topology snippets.
tools/topology/topology2/sof-hda-generic-wov-manifest.conf	New manifest-style topology source including WoV/VAD/KPB assets.
tools/topology/topology2/sof-hda-generic-manifest.conf	New manifest-style topology source for HDA generic.
tools/topology/topology2/sof-hda-generic-4ch.conf	New manifest-style topology source variant.
tools/topology/topology2/platform/intel/dmic-wov-multi.conf	New multi-detector WoV topology definition (DMIC multi-slot).
tools/topology/topology2/platform/intel/dmic-generic.conf	Switches/extends DMIC includes to pull in the new multi-WoV topology.
tools/topology/topology2/include/pipelines/cavs/wov-detect.conf	Adjusts pipeline class formatting metadata.
tools/topology/topology2/include/pipelines/cavs/dai-kpb-be.conf	Modifies the DAI→KPB pipeline class defaults and routing source.
tools/topology/topology2/include/components/wov.conf	Changes WoV control naming to be index-qualified.
tools/topology/topology2/include/components/wov-arbiter.conf	New Topology2 widget class for the arbiter component.
tools/topology/topology2/include/components/vad-gate.conf	New Topology2 widget class for the VAD gate component.
tools/topology/topology2/include/components/kpb.conf	Switches KPB widget UUID to the `kpb4` static module UUID.
tools/topology/topology2/dmic-wov-multi-manifest.conf	New manifest-style topology for the multi-slot WoV design.
tools/rimage/config/tgl.toml.h	Updates rimage TOML include set for new modules.
src/samples/audio/detect_test.toml	Increases instance count for the detector sample module.
src/samples/audio/detect_test.c	Extends detect_test with slot awareness, AMS integration, and new detection behavior.
src/include/sof/lib/ams_msg.h	Adds new AMS UUIDs and payload structs for arbiter↔detector control.
src/include/sof/audio/wov_arbiter.h	New public header for arbiter constants/IPC param IDs.
src/include/sof/audio/vad_gate.h	New public header for VAD gate runtime config/IPC param IDs.
src/include/sof/audio/kpb.h	Adds a new detector lookup prototype.
src/include/ipc4/detect_test.h	Adds IPC4 large-config param for setting WoV slot ID.
src/audio/wov_arbiter/wov_arbiter.toml	New module metadata for rimage/LLEXT.
src/audio/wov_arbiter/wov_arbiter.c	New arbiter component implementation.
src/audio/wov_arbiter/README.md	New architecture and testing documentation for multi-slot WoV.
src/audio/wov_arbiter/CMakeLists.txt	Builds the new arbiter component.
src/audio/vad_gate/vad_gate.toml	New module metadata for rimage/LLEXT.
src/audio/vad_gate/vad_gate.c	New VAD gate component implementation.
src/audio/vad_gate/Kconfig	Kconfig option for building the VAD gate component.
src/audio/vad_gate/CMakeLists.txt	Builds the new VAD gate component.
src/audio/mixin_mixout/mixin_mixout.c	Modifies scheduling/state/copy behavior in mixin/mixout processing.
src/audio/kpb.toml	Increases instance count for KPB module.
src/audio/kpb.c	Modifies KPB binding/prepare/copy behavior and downstream triggering.
src/audio/Kconfig	Adds Kconfig entry for the new arbiter component and sources VAD gate Kconfig.
src/audio/CMakeLists.txt	Adds build integration for `vad_gate` and `wov_arbiter`.
app/prj.conf	Enables samples / sample keyphrase in the app config.
app/debug_overlay.conf	Enables fatal breadcrumb recording in debug overlay.
app/boards/intel_adsp_cavs25.conf	Enables new components and sample config on the cavs25 board.

Suppressed comments (2)

src/audio/mixin_mixout/mixin_mixout.c:506

Calling comp_copy() on other components from inside mixin_process() is likely to cause re-entrancy/recursion and breaks the normal scheduler-driven copy order. This should be removed and left to the pipeline scheduler.

 if (mixout_mod && mixout_mod->dev)
comp_copy(mixout_mod->dev);

src/audio/mixin_mixout/mixin_mixout.c:607

mixout_process() manually triggers comp_copy() on the downstream sink component (e.g., KPB). This can cause recursive copy calls and violates the scheduler’s ownership of graph execution order.

 comp_err(dev, "mixout_process: sink_buf=0x%x produced=%u bytes, triggering kpb=0x%x",
sink_buf, bytes_to_produce, dev_comp_id(sink_comp));
comp_copy(sink_comp);

 e93326d8-0d14-4bf0-bcb9e063d3d80136 twb_sched
 42f8060c-832f-4dbf-b24751e961997b34 up_down_mixer
 6f6b6f4b-6f73-7466-20e1e62b9779f003 userspace_proxy
+8c7d6e5f-4a3b-2c1d-0e9f8a7b6c5d4e3f vad_gate


 1028070e-04e8-46ab-8d8110a0116ce738 wait
 d944281a-afe9-4695-a043d7f62b89538e waves
 13c8bc59-c4fa-4ad1-b93ace97cd30acc7 wdt
+4a5b6c7d-8e9f-4a1b-2c3d4e5f60718293 wov_arbiter


 #endif

-#ifdef CONFIG_SAMPLE_KEYPHRASE
 #include <samples/audio/detect_test.toml>


 <vendor-token.conf>
 <tokens.conf>
 <virtual.conf>
+<vad-gate.conf>


+	KWD_TEST_UUID "1f:d5:a8:eb:27:78:b5:47:ee:82:de:6e:77:43:af:67"
+	WOV_UUID "1f:d5:a8:eb:27:78:b5:47:ee:82:de:6e:77:43:af:67"


+ if (mixout->state != COMP_STATE_ACTIVE)
+ mixout->state = COMP_STATE_ACTIVE;



+ struct comp_dev *wov_comp = sink ? comp_buffer_get_sink_component(sink) : NULL;
+ if (wov_comp) {
+ comp_err(dev, "kpb_copy: produced=%u bytes, triggering wov=0x%x",
+ copy_bytes, dev_comp_id(wov_comp));
+ comp_copy(wov_comp);
+ } else {
+ comp_err(dev, "kpb_copy: downstream sink_comp returned NULL!");
+ }


+}

-LOG_MODULE_REGISTER(kd_test, CONFIG_SOF_LOG_LEVEL);
+LOG_MODULE_REGISTER(kd_test, LOG_LEVEL_INF);


+	comp_err(dev, "notify_host: WOV module_id=0x%x instance_id=0x%x slot_id=%u detected",
+ dev_comp_id(dev) >> 16, dev_comp_id(dev) & 0xffff, cd->wov_slot_id);


 static const struct comp_driver comp_keyword;
+static struct comp_dev *wov_detect_devs[256];
+
+struct comp_dev *get_wov_detector_comp(uint32_t ppl_id)
+{
+	if (ppl_id < 256)
+ return wov_detect_devs[ppl_id];
+	return NULL;
+}


kv2019i

Some comments inline

kv2019i · 2026-08-06T12:11:10Z

- continue;
- }
+ if (mixout->state != COMP_STATE_ACTIVE)
+ mixout->state = COMP_STATE_ACTIVE;


This seems to revert existing functionality that on purpose skipped non-active mixouts. The removed comment states non-active mixouts might not be configured, making this even more unsafe.

kv2019i · 2026-08-06T12:14:07Z

+	sink = comp_dev_get_first_data_consumer(dev);
+
+	frame_bytes = audio_stream_frame_bytes(&source->stream);
+	frames = audio_stream_get_avail_bytes(&source->stream) / frame_bytes;


Would be good to se the new pipeline2.0 style source/sink interface from the start (like the flurry of recent PRs to convert existing components).

kv2019i · 2026-08-06T12:15:01Z

@@ -0,0 +1,344 @@
+// SPDX-License-Identifier: BSD-3-Clause
+//
+// Copyright(c) 2024 Intel Corporation. All rights reserved.


Copyright statement needs update

kv2019i · 2026-08-06T12:19:47Z

+ const int32_t *src = audio_stream_read_frag_s32(s, i);
+ int32_t diff = abs(*src) - abs(cd->energy);
+ cd->energy += diff >> cd->config.energy_shift;
+	}


This is quite simple (albeit very cheap to compute). We do have a bit more logic in MFCC VAD with noise floor tracking already (#10782 ).

kv2019i · 2026-08-06T12:23:13Z

@@ -0,0 +1,455 @@
+// SPDX-License-Identifier: BSD-3-Clause
+//
+// Copyright(c) 2024 Intel Corporation. All rights reserved.


kv2019i · 2026-08-06T12:25:20Z

+ */
+	uint8_t active_slot;
+
+#if CONFIG_AMS


Not related to this PR, but I'm wondering whether we should/need to depend on AMS. It is currently disabled in many Intel configs.

naveen-manohar · 2026-08-10T08:14:52Z

+ }
+	}
+
+	kpb_change_state(kpb, KPB_STATE_RUN);


kpb enters run too early ? another run L1012
should we check after all sink prepare & its check completed ?

naveen-manohar · 2026-08-10T08:48:39Z

+}

-LOG_MODULE_REGISTER(kd_test, CONFIG_SOF_LOG_LEVEL);
+LOG_MODULE_REGISTER(kd_test, LOG_LEVEL_INF);


switch to CONFIG_SOF_LOG_LEVEL for consistency ?

lgirdwood

please comment every new code block

lgirdwood · 2026-08-11T13:09:59Z

+ sink_params.sample_container_bytes = 4;
+ sink_params.sample_valid_bytes = 4;
+ sink_params.frame_fmt = SOF_IPC_FRAME_S32_LE;


these should come from the topology for kpb object

lgirdwood · 2026-08-11T13:12:29Z

+ if (sink_comp && sink_comp->state == COMP_STATE_INIT) {
+ struct sof_ipc_stream_params sink_params;
+ memset_s(&sink_params, sizeof(sink_params), 0, sizeof(sink_params));
+ sink_params.channels = kpb->config.channels ? kpb->config.channels : 2;
+ sink_params.rate = kpb->config.sampling_freq ? kpb->config.sampling_freq : 16000;
+ sink_params.sample_container_bytes = 4;
+ sink_params.sample_valid_bytes = 4;
+ sink_params.frame_fmt = SOF_IPC_FRAME_S32_LE;
+ comp_params(sink_comp, &sink_params);
+ comp_prepare(sink_comp);


this looks like teh same code block as line 992, we should make this a statice line if its doing teh same thing as 992, cooment it and also get config data from topology rather than hard code.

lgirdwood · 2026-08-11T13:14:36Z

- ret = -EBUSY;
+ if (kpb->host_sink) {
+ kpb_change_state(kpb, KPB_STATE_RESETTING);
+ ret = -EBUSY;


can we inline comment when reset will be done since we break here.

lgirdwood · 2026-08-11T14:08:31Z

- /* Discard data if sink is not active */
- if (comp_buffer_get_sink_component(sink)->state != COMP_STATE_ACTIVE) {
- copy_bytes = audio_stream_get_avail_bytes(&source->stream);
- comp_update_buffer_consume(source, copy_bytes);
- comp_dbg(dev, "KD not active, dropping %zu bytes...", copy_bytes);
- break;
- }


why do we need to drop this, if downstream is not active then this seems like the right thing todo.

lgirdwood · 2026-08-11T14:09:37Z

-	kpb_clear_history_buffer(buff);
+


why do we no longer need to clear ?

lgirdwood · 2026-08-11T14:19:21Z

+#define AMS_WOV_DETECT_MSG_UUID { 0xc3, 0xd7, 0xe8, 0x41, 0x12, 0xf0, 0x4e, \
+ 0x8a, 0xb9, 0x01, 0x5a, 0x6b, 0x7c, 0x8d, \
+ 0x9e, 0x0f }
+
+#define AMS_WOV_CTRL_MSG_UUID { 0xf1, 0xe2, 0xd3, 0xc4, 0xb5, 0xa6, 0x47, \
+ 0x89, 0x8a, 0xce, 0x12, 0x34, 0x56, 0x78, \
+ 0x90, 0xab }


what are the Zephyr options here instead of AMS ? i.e. what maps at feature level to AMS and also works well in userspace context

lgirdwood · 2026-08-11T14:20:36Z

+#if CONFIG_IPC_MAJOR_4
+	struct ipc4_base_module_cfg base_cfg;
+#endif


lets make this IPC4 only via Kconfig and remove the ifdefs

lgirdwood · 2026-08-11T14:26:15Z

+# to three KPB instances each fronted by a detect_test WOV detector.
+# When any detector fires, the WOV arbiter drains that KPB to the host PCM
+# capture stream and pauses the remaining two via AMS.
+#
+# Pipeline IDs
+# 100 DAI pipeline : HDA copier → VAD Gate → Mixin
+# 101 KPB pipeline : Mixout_0 → KPB_0 → WOV_0 (Core 0)


These need to be "feature" topologies and they should be able to use any capture endpoint e.g. SDW DMIC, PCM, DMIC, I2S and HDA Mics

lgirdwood · 2026-08-11T14:29:22Z

+
+After `arecord` opens `hw:0,11` (pipeline prepare):
+
+```
+kd_test.test_keyword_prepare: comp:4 0x2000d kd_dp thread started for slot 2 (pinned to core 1)


can we include example command to generate test tones that can be used to test the WOV.

lgirdwood · 2026-08-11T14:29:55Z

+running on the DSP. Each detector has its own Keyphrase Buffer (KPB) that continuously records
+a pre-roll window (6 seconds on TigerLake, 2.1 seconds on other platforms). When any detector


we just have one KPB now ?

lgirdwood · 2026-08-11T14:32:00Z

Please also add a enum kcontrol that will be RO and volatile and will reflect the ID of the WOV that triggered the wake word.

Allow multiple downstream WOV detector pipelines to share one KPB instance by extending the KPB to support a sel_sink drain path alongside the existing dedicated host_sink path. Key changes: - Add sel_sink field to kpb_data to track the downstream WOV detector sink; kpb_set_sink() assigns it from the component bind call. - kpb_init_draining: when host_sink is NULL (multi-KPB WOV topology with no dedicated PCM capture), redirect the pre-roll drain through sel_sink so that history reaches the wov_arbiter and onward to the host copier. Initialise host_period_size from sel_sink stream geometry when not already set. Skip pausing the selector component when sel_sink is the active drain path. - kpb_init_draining: cap drain_req to the actual buffered amount instead of aborting when less history is available than requested (partial pre-roll is better than none). - kpb_reset: add immediate-reset path when host_sink==NULL in BUFFERING/DRAINING state; the LL scheduler is gone after STOP so the async-EBUSY path can never complete. - kpb_init_draining: guard against NULL host_sink to prevent NULL deref on WOV-only instances. - Fix fallback sink assignment in kpb_copy: use && instead of || so the fallback only fires when BOTH sel_sink and host_sink are NULL. - Raise KPB_MAX_BUFF_TIME to 6000 ms; clear HOST_WAKEUP_TIME (no extra delay needed with direct sel_sink routing). - kpb.conf: expose host_sink_index parameter for topology binding. - Reduce log noise: demote two comp_err to comp_dbg in the RUN copy path. Signed-off-by: Liam Girdwood <liam.r.girdwood@linux.intel.com>

Add a lightweight VAD gate component that sits between the DAI copier and the mixin fan-out in the WOV capture pipeline. It measures per-frame signal energy and suppresses propagation of near-silence frames to the downstream KPB/detector branches, preventing unnecessary wakeups and reducing idle CPU load. The gate passes frames when the mean absolute sample value exceeds a configurable threshold (VAD_THRESHOLD, default 0 = bypass for testing). Suppressed frames are replaced with PPL_STATUS_PATH_STOP so the LL scheduler can idle the downstream pipelines. Kconfig: CONFIG_COMP_VAD_GATE (depends on IPC_MAJOR_4). Topology: vad-gate.conf widget class for topology2. Signed-off-by: Liam Girdwood <liam.r.girdwood@linux.intel.com>

Add two AMS message UUIDs and their payload structures for the multi-slot WOV arbitration protocol: AMS_WOV_DETECT_MSG_UUID — sent from a keyword detector to the wov_arbiter when a keyword is confirmed. Payload: wov_detect_payload carrying the slot_id (0..N-1) of the detector that fired. AMS_WOV_CTRL_MSG_UUID — sent from the wov_arbiter to all registered keyword detectors. Payload: wov_ctrl_payload carrying a cmd field (WOV_CTRL_CMD_PAUSE or WOV_CTRL_CMD_RESUME) and the active_slot (WOV_ARB_NO_ACTIVE=0xff on RESUME to address all slots). Signed-off-by: Liam Girdwood <liam.r.girdwood@linux.intel.com>

Add the wov_arbiter audio component. It receives audio from up to WOV_ARB_MAX_SLOTS (3) KPB host-sink pins and routes exactly one slot's audio to the single host-copier downstream of it. Before any keyword fires the arbiter fills the host sink with silence (memset zero). When any detector sends an AMS_WOV_DETECT message the arbiter activates that slot (first-wins), broadcasts AMS_WOV_CTRL PAUSE to all other detectors, and starts forwarding KPB pre-roll + live audio to the host. On stream STOP or PAUSE the arbiter resets to NO_ACTIVE and broadcasts RESUME so all detectors restart listening. IPC4 debug parameter param_id=1 (IPC4_WOV_ARB_SET_ACTIVE_SLOT) allows forcing a slot active without AMS, useful for lab testing. The IBS/OBS is set to 32 bytes (matching the 1 ms DMIC LL period). The arbiter fills exactly sink_free bytes on every copy to avoid host copier underruns during the pre-detection silence phase. Kconfig: CONFIG_COMP_WOV_ARBITER (auto-selects CONFIG_AMS). UUID: 4a5b6c7d-8e9f-4a1b-2c3d-4e5f60718293. Topology: wov-arbiter.conf widget class for topology2. Signed-off-by: Liam Girdwood <liam.r.girdwood@linux.intel.com>

Enable CONFIG_COMP_WOV_ARBITER, CONFIG_COMP_VAD_GATE, CONFIG_COMP_KPB, CONFIG_SAMPLES, and CONFIG_SAMPLE_KEYPHRASE on the intel_adsp_cavs25 board (TigerLake, CAVS2.5) to include the multi-slot WOV pipeline in default firmware builds for that platform. Signed-off-by: Liam Girdwood <liam.r.girdwood@linux.intel.com>

Extend detect_test (the sample keyword detector) to participate in the multi-slot WOV arbitration system coordinated by wov_arbiter. Slot assignment: A static wov_slot_id is derived from the pipeline ID at component creation time (pipeline 101/1 → slot 0, 102/3 → slot 1, 103/4 → slot 2). The slot ID can be overridden via IPC4 LARGE_CONFIG_SET param_id=4 (IPC4_DETECT_TEST_SET_WOV_SLOT). DP thread batching: Each slot runs its own Zephyr k_thread at K_PRIO_PREEMPT(12) with a 4096-byte stack. The LL copy path accumulates 320-frame (20 ms at 16 kHz) S16_LE samples into a double-buffer and gives a semaphore when a batch is ready. The DP thread wakes, runs the detection algorithm on the completed buffer, and signals the LL thread to switch sides. Slot 2 is additionally pinned to DSP Core 1 via k_thread_cpu_pin(). Threads are started in prepare() and stopped in reset() + free(). AMS integration: On detection, detect_test_notify() sends: 1. SOF_IPC4_NOTIFY_PHRASE_DETECTED IPC4 notification to the host (word_id = wov_slot_id). 2. AMS_KPD_MSG to the KPB to initiate pre-roll drain on host_sink. 3. AMS_WOV_DETECT_MSG to the wov_arbiter. The component registers as a consumer of AMS_WOV_CTRL_MSG and responds to PAUSE (stops detecting) and RESUME (resets cd->detected, resumes). Rotating auto-trigger (test mode): When CONFIG_AMS is enabled a per-slot auto-trigger fires after the first 320-frame DP batch if this slot matches the current session target (kd_test_session % KD_MAX_SLOTS). kd_test_session is incremented in reset() when slot 0 resets, so each arecord session cycles through slots 0 → 1 → 2 → 0 without a firmware reload. Audio pass-through: detect_test.copy() forwards source audio to the downstream wov-arbiter sink so the real-time data path remains live both before and after detection. Signed-off-by: Liam Girdwood <liam.r.girdwood@linux.intel.com>

Add a standalone three-slot WOV topology (dmic-wov-multi.conf / dmic-wov-multi-manifest.conf) driven by a single DMIC input with: Pipeline 100 (Core 0): DAI Copier → vad_gate → mixin (1→3 fan-out) Pipeline 101 (Core 0): mixout → kpb → detect_test (slot 0, Male) Pipeline 102 (Core 0): mixout → kpb → detect_test (slot 1, Female) Pipeline 103 (Core 1): mixout → kpb → detect_test (slot 2, Child) Pipeline 104 (Core 0): wov_arbiter → host-copier (PCM 11, hw:0,11) Each KPB has two output pins: sel_sink (to the detector) and host_sink (to the arbiter). When a detector fires, it switches KPB to drain via host_sink; the arbiter routes that slot's audio to the host copier. Also included: sof-hda-generic-wov-manifest.conf — build entry point for the WOV topology variant (includes dmic-wov-multi.conf via INCLUDE_WOV). Bug fixes to shared topology2 infrastructure: wov.conf: rename wov_init → wov_init_$index so multiple wov widget instances do not share a control name. dai-kpb-be.conf: add dai_type/copier_type/stream_name fields to the dai-copier widget; fix channel config for 4-channel DMIC; source route now uses $DMIC_WOV_DAI_PIPELINE_SRC for flexibility. wov-detect.conf: add num_input/output_audio_formats=1 to the micsel widget (mandatory attribute missing from the original class). sof-hda-generic.conf: add WOV widget class includes (dai-kpb-be, wov-detect, vad-gate, kpb, wov, pin-binding helpers) so the generic HDA topology can instantiate WOV pipelines when INCLUDE_WOV is set; remove duplicate vad-gate.conf include. tgl.toml.h: restore CONFIG_SAMPLE_KEYPHRASE guard on detect_test.toml; add CONFIG_COMP_VAD_GATE and CONFIG_COMP_WOV_ARBITER guards on the new vad_gate/wov_arbiter toml includes. Compile with: alsatplg -I tools/topology/topology2 -p \ -c tools/topology/topology2/dmic-wov-multi-manifest.conf \ -o sof-hda-generic-wov.tplg Signed-off-by: Liam Girdwood <liam.r.girdwood@linux.intel.com>

Document the multi-slot WOV subsystem comprehensively: Architecture: - Component graph mermaid diagram showing all 5 pipelines (100-104), their component chains, and the AMS message flows. - Key design points table (format, pre-roll duration, slot count, arbitration policy, core affinity, PCM device). Signal processing flow: - LL thread flowchart: HW DMA → DAI → VAD gate → mixin → mixout → KPB → detect_test accumulation → k_sem_give. - DP thread flowchart: semaphore wake → algorithm → notify or auto-trigger (test mode). - KPB drain sequence: AMS → sel_sink switch → ring drain → live audio. Arbiter state machine: - stateDiagram-v2 with Idle/Active transitions and first-wins guard. - Routing logic table (silence vs active slot vs idle slot output). AMS inter-module messaging: - Message catalog with UUID constants, directions, payload structs. - Full detect-to-drain sequenceDiagram from DMIC capture to host close. Firmware API reference: - IPC4 LARGE_CONFIG_SET parameter tables for detect_test and wov_arbiter. - AMS producer/consumer registration code examples. - detect_test_notify() three-step contract. - DP thread ping-pong buffer field semantics. Linux host API: - ALSA PCM capture (hw:0,11) via arecord and libasound. - snd_ctl voice detection notification polling in userspace. - Capture timing model (silence → pre-roll → live audio transition). - IPC4 slot override via sof-ctl. Adding a new WOV algorithm: - Approach A: modify detect_test.c directly (prototype path). - Approach B: new native SOF module with full skeleton (AMS reg, IPC4 new/prepare/copy/notify, Kconfig, CMakeLists, UUID registration). - Approach C: IADK/LLEXT loadable binary for third-party IP. Topology: build and deploy guide: - Source layout, alsatplg compile command, generic deploy instructions. - Firmware build and deploy (generic build dir, generic DUT SCP). - Topology config reference table. Adding a fourth slot. Build system: Kconfig minimum set, module UUID catalog. Testing and verification: - Three-slot rotating auto-trigger test with Python RMS verification. - Real-audio frequency sweep test. - Firmware trace (mtrace) reading procedure and expected log events. Signed-off-by: Liam Girdwood <liam.r.girdwood@linux.intel.com>

Expose the new dmic-wov-multi topology through two non-invasive hooks in dmic-generic.conf, without disturbing existing single-slot WOV users: INCLUDE_WOV "multi": includes dmic-wov-multi.conf, selecting the 3-slot KPB/arbiter topology. The existing "true" value continues to select the original dmic-wov.conf (single-slot, unchanged). DMIC1_ENABLE "wov-multi": includes dmic-wov-multi.conf as an alternative DMIC1 pipeline mode, alongside the existing "passthrough" and "mfcc" values. Build the multi-slot WOV topology variant with: alsatplg ... -DINCLUDE_WOV=multi Signed-off-by: Liam Girdwood <liam.r.girdwood@linux.intel.com>

lgirdwood

Needs all review comments to be addressed, also plan AMS alternative.

lgirdwood · 2026-08-13T11:52:04Z

+	/* WOV topology has exactly 2 output pins:
+ * pin 0 (buf_id=0) -> sel_sink (WOV detector)
+ * pin 1 (buf_id=IPC4_COMP_ID(1,0)=0x10) -> host_sink (arbiter drain) */
+	comp_dev_for_each_consumer(dev, sink) {
+ if (buf_get_id(sink) == 0)
+ kpb->sel_sink = sink;
+ else if (!kpb->host_sink)
+ kpb->host_sink = sink;
+ else
+ comp_warn(dev, "kpb_prepare: unexpected 3rd consumer pin, buf_id=0x%x",
+ buf_get_id(sink));
+	}


If these are always hard coded, we should really use enums for the buf_id and I expect the host and sel sinks should be set from topology via IPC4 bind calls ?

lgirdwood · 2026-08-13T11:52:49Z

+ sink_params.sample_container_bytes = 4;
+ sink_params.sample_valid_bytes = 4;
+ sink_params.frame_fmt = SOF_IPC_FRAME_S32_LE;


these 3 are hard coded, they should come from topology ?

lgirdwood · 2026-08-13T11:56:17Z

+#define VAD_DEFAULT_THRESHOLD 0 /* bypass — tone-friendly for lab testing */
+#define VAD_DEFAULT_ONSET_FRAMES 3 /* frames above threshold before SPEECH */
+#define VAD_DEFAULT_HANGOVER 30 /* frames below threshold before SILENCE */
+#define VAD_DEFAULT_ENERGY_SHIFT 6 /* IIR alpha = 1/2^6 */


should come from topology.

lgirdwood requested review from abonislawski, naveen-manohar and udaymb July 22, 2026 12:07

lgirdwood commented Jul 22, 2026
View reviewed changes

Comment threadsrc/audio/wov_arbiter/wov_arbiter.c Outdated
Comment threadsrc/audio/wov_arbiter/wov_arbiter.c
Comment threadsrc/audio/CMakeLists.txt Outdated
Comment threadsrc/audio/Kconfig Outdated
Comment threadsrc/include/sof/audio/wov_arbiter.h Outdated
Comment threaduuid-registry.txt Outdated

lgirdwood force-pushed the feature/wov-multi-kpb branch from 49f3dc6 to c3c7161Compare July 22, 2026 13:13

lgirdwood changed the title ~~WIP: audio/topology: add multi-KPB WOV arbiter for 3-keyword DMIC capture~~WIP: audio/topology: add multi-KPB WOV arbiter for 3-keyword DMIC capture with VADJul 23, 2026

TsaiGaggery mentioned this pull request Jul 23, 2026
audio: iadk: support API 4.5.2 loadable detector modules #11032
Closed

lgirdwood force-pushed the feature/wov-multi-kpb branch 2 times, most recently from 0f6f889 to 3af37c0Compare July 29, 2026 15:22

lgirdwood marked this pull request as ready for review July 29, 2026 15:23

lgirdwood requested review from dbaluta, fkwasowi, iganakov, jsarha, kv2019i, lbetlej, mmaka1, pblaszko, plbossart, ranj063 and tmleman as code owners July 29, 2026 15:23

lgirdwood changed the title ~~WIP: audio/topology: add multi-KPB WOV arbiter for 3-keyword DMIC capture with VAD~~kpb: add multi-KPB WOV arbiter for 3-keyword DMIC capture with VADJul 29, 2026

lgirdwood commented Jul 29, 2026
View reviewed changes

lyakh requested changes Jul 30, 2026
View reviewed changes

lgirdwood requested a review from Copilot August 1, 2026 12:38

Copilotstarted reviewing on behalf of lgirdwood August 1, 2026 12:39 View session

CopilotAI reviewed Aug 1, 2026
View reviewed changes

kv2019i requested changes Aug 6, 2026
View reviewed changes

lgirdwood force-pushed the feature/wov-multi-kpb branch from 3af37c0 to a79115bCompare August 9, 2026 11:12

naveen-manohar reviewed Aug 10, 2026
View reviewed changes

lgirdwood force-pushed the feature/wov-multi-kpb branch from a79115b to f45334fCompare August 11, 2026 13:04

lgirdwood commented Aug 11, 2026
View reviewed changes

lgirdwood marked this pull request as draft August 11, 2026 14:31

lgirdwood force-pushed the feature/wov-multi-kpb branch from f45334f to 9458dbdCompare August 11, 2026 17:34

lrgirdwo added 9 commits August 12, 2026 17:32

lgirdwood force-pushed the feature/wov-multi-kpb branch from 9458dbd to a59bf45Compare August 12, 2026 17:05

lgirdwood commented Aug 13, 2026
View reviewed changes

		KWD_TEST_UUID "1f:d5:a8:eb:27:78:b5:47:ee:82:de:6e:77:43:af:67"
		WOV_UUID "1f:d5:a8:eb:27:78:b5:47:ee:82:de:6e:77:43:af:67"

		if (mixout->state != COMP_STATE_ACTIVE)
		mixout->state = COMP_STATE_ACTIVE;

		comp_err(dev, "notify_host: WOV module_id=0x%x instance_id=0x%x slot_id=%u detected",
		dev_comp_id(dev) >> 16, dev_comp_id(dev) & 0xffff, cd->wov_slot_id);

		running on the DSP. Each detector has its own Keyphrase Buffer (KPB) that continuously records
		a pre-roll window (6 seconds on TigerLake, 2.1 seconds on other platforms). When any detector

Uh oh!

Conversation

lgirdwood commented Jul 22, 2026• edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

lgirdwood left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

lgirdwood left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

CopilotAI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

kv2019i left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

lgirdwood commented Jul 22, 2026•
edited
Loading