ABX00087 Boto ea Nts'etsopele ea WiFi ea UNO R4
Cricket Shot Recognition e sebelisa Arduino UNO R4 WiFi + ADXL345 + Edge
Tšusumetso
Tokomane ena e fana ka ts'ebetso e felletseng ea ho aha sistimi ea ho tseba batho ba krikete ka Arduino UNO R4 WiFi e nang le ADXL345 accelerometer le Edge Impulse Studio. Morero o kenyelletsa ho bokella lintlha tsa accelerometer, ho koetlisa mohlala oa ho ithuta mochini, le ho khutlisetsa mohlala o koetlisitsoeng ho Arduino bakeng sa ho arola lithunya tsa nako ea nnete.
Lits'oants'o tsa krikete li nahanoa morerong ona:
– Koahela Drive
– Straight Drive
– Hula Shot
Mohato oa 1: Litlhoko tsa Hardware
- Arduino UNO R4 WiFi
– ADXL345 Accelerometer (I2C)
– Lithapo tsa jumper
- Breadboard (ho ikhethela)
– thapo ya USB Type-C
Mohato oa 2: Litlhoko tsa Software
- Arduino IDE (ea morao tjena)
- Ak'haonte ea Edge Impulse Studio (mahala)
- Lisebelisoa tsa Edge Impulse CLI (Node.js ea hlokahala)
- laebrari ea Adafruit ADXL345
Mohato oa 3: Wiring ADXL345
Hokela sensor ea ADXL345 ho Arduino UNO R4 WiFi ka tsela e latelang:
VCC → 3.3V
GND → GND
SDA → SDA (A4)
SCL → SCL (A5)
CS → 3.3V (boikhethelo, bakeng sa mokhoa oa I2C)
SDO → phaphametseng kapa GND
Mohato oa 4: Lokisa Sensor ea IDE
U ka kenya Lilaebrari tsa Sensor joang ho Arduino IDE?
Bula Arduino IDE
Bula Lisebelisoa → Laola Lilaebrari… 'me u kenye: Adafruit ADXL345 Unified Adafruit Unified Sensor
(Haeba u na le LSM6DSO kapa MPU6050 sebakeng seo: kenya SparkFun LSM6DSO , Adafruit LSM6DS kapa MPU6050 ka ho loketseng.)
Mohato oa 5: Sketch ea Arduino bakeng sa Pokello ea Lintlha
Kenya setšoantšo sena ho Arduino UNO R4 WiFi ea hau. E tsamaisa data ea accelerometer ka sebopeho sa CSV (x,y,z) ho ~ 18 Hz bakeng sa Edge Impulse.
# kenyeletsa
# kenyeletsa
Adafruit_ADXL345_Unified accel =
Adafruit_ADXL345_Unified(12345);
ho seta letho () {
Serial.qala(115200);
haeba (!accel.begin()) {
Serial.println("Ha ho ADXL345 e fumanoeng");
ha (1);
}
accel.setRange(ADXL345_RANGE_4_G);
}
loop e se nang letho () {
sensor_event e;
accel.getEvent(&e);
Seri.print (e.acceleration.x);
Seri.print(“,”);
Seri.print(e.acceleration.y);
Seri.print(“,”);
Serial.println(e.acceleration.z);tieho(55); // ~ 18 Hz
}
Beha Edge Impulse
Mohato oa 6: Ho hokela ho Edge Impulse
- Koala Arduino Serial Monitor.
- Sebelisa taelo: Edge-impulse-data-forwarder -frequency 18
- Kenya mabitso a axis: accX, accY, accZ
- Bolela sesebelisoa sa hau: Arduino-Cricket-Board
- Netefatsa khokahano ho Edge Impulse Studio tlasa 'Lisebelisoa'.
Mohato oa 7: Pokello ea Lintlha
Ho Edge Impulse Studio → Ho fumana lintlha:
- Sesebelisoa: Arduino-Cricket-Boto
- Sensor: Accelerometer (3 lilepe)
– Sampbolelele: 2000 ms (2 metsotsoana)
- Khafetsa: 18 Hz
Ngola bonyane 40 samphanyane ka sehlopha:
– Koahela Drive
– Straight Drive
– Hula Shot
Bokella Lintlha Examples
Koahela Drive
Sesebelisoa: Arduino-Cricket-Boto
Label: Koahela Drive
Sensor: Sensor e nang le lilepe tse 3 (accX, accY, accZ)
Sampbolelele: 10000ms
Khafetsa: 18 Hz
Example Raw Data:
accX -0.32
accY 9.61
accZ -0.12
Otlolohile Drive
Sesebelisoa: Arduino-Cricket-Boto
Label: Straight Drive
Sensor: Sensor e nang le lilepe tse 3 (accX, accY, accZ)
Sampbolelele: 10000ms
Khafetsa: 18 Hz
Example Raw Data:
accX 1.24
accY 8.93
accZ -0.42
Hula Thunya
Sesebelisoa: Arduino-Cricket-Boto
Label: Hula Shot
Sensor: Sensor e nang le lilepe tse 3 (accX, accY, accZ)
Sampbolelele: 10000 ms
Khafetsa: 18 Hz
Example Raw Data:
accX 2.01
accY 7.84
accZ -0.63 
Mohato oa 8: Moralo oa Impulse
Bula Theha tšusumetso:
Thibelo ea ho kenya: Lintlha tsa letoto la nako (li-axes tse 3).
Boholo ba fensetere: 1000 ms Keketseho ea fensetere (mohato): 200 ms Nobles: Lilepe, Boholo (boikhethelo), maqhubu a 18.
Thibelo ea ho sebetsa: Tlhahlobo ea Spectral (aka Spectral Features bakeng sa ho sisinyeha). Boholo ba fensetere: 1000 ms Keketseho ea lifensetere (mohato): 200 ms Nobles: Lilepe, Boholo (boikhethelo), boloka tsohle tse behiloeng pele.
Sebaka sa ho ithuta: Sehlopha (Keras).
Tobetsa Boloka tšusumetso. 
Hlahisa likarolo:
Eya ho Spectral analysis, tobetsa Save parameters, ebe Hlahisa likarolo tsa sete ea koetliso.
Koetlisa mohlala o monyenyane
E-ea ho Classifier (Keras) 'me u sebelise compact config joaloka:
Neural network: 1-2 lihlopha tse teteaneng (mohlala, 60 → 30), ReLU
Nako: 40-60
Sekhahla sa ho ithuta: 0.001-0.005
Boholo ba sehlopha: 32
Karohano ea data: 80/20 (terene / teko)
Boloka le ho koetlisa lintlha
Lekola 'me u hlahlobe tlhahlobo ea Model ka sete ea ho boloka.
Hlahloba matrix a pherekano; haeba selikalikoe le holimo li kopana, bokella lintlha tse fapaneng tse fapaneng kapa tweak
Litekanyetso tsa Spectral (boholo ba fensetere / mokatong oa lerata).
Mohato oa 9: Ho fetisetsoa ho Arduino
Eya ho Deployment:
Khetha laebrari ea Arduino (laebrari ea C ++ le eona ea sebetsa).
Numella EON Compiler (haeba e le teng) ho fokotsa boholo ba mohlala. 
Khoasolla .zip, ebe ho Arduino IDE: Sketch → Kenyelletsa Library → Eketsa .ZIP Library… Sena se eketsa exampjoalo ka Static buffer le Continuous under File → Exampjoalo →
Lebitso la Morero oa hau - Edge Impulse. Sekeche sa inference bakeng sa Arduino UNO EK R4 WiFi + ADXL345.
Mohato oa 10: Sketch ea Inference ea Arduino
# kenyeletsa
# kenyeletsa
# kenyeletsa // Kenya sebaka ka hlooho ea Edge Impulse
Adafruit_ADXL345_Unified accel =
Adafruit_ADXL345_Unified(12345);
static bool debug_nn = bohata;
ho seta letho () {
Serial.qala(115200);
ha (!Seriale) {}
haeba (!accel.begin()) {
Serial.println(“ERROR: ADXL345 ha e-so fumanwe”);
ha (1);
}
accel.setRange(ADXL345_RANGE_4_G);
}
loop e se nang letho () {
float buffer[EI_CLASSIFIER_DSP_INPUT_FRAME_SIZE] = {0};
bakeng sa (size_t ix = 0; ix < EI_CLASSIFIER_DSP_INPUT_FRAME_SIZE; ix +=
3) {
uint64_t next_tick = micros() + (EI_CLASSIFIER_INTERVAL_MS *
1000);
sensor_event e;
accel.getEvent(&e);
buffer[ix + 0] = e.acceleration.x;
buffer[ix + 1] = e.acceleration.y;
buffer[ix + 2] = e.acceleration.z;
int32_t wait = (int32_t)(next_tick – micros());
haeba (emela > 0) liehaMetsotsoana e meng(emela);
}
lets'oao la_t;
int err = numpy::signal_from_buffer(buffer,
EI_CLASSIFIER_DSP_INPUT_FRAME_SIZE, &signal);
haeba (phoso != 0) khutla;
ei_impulse_result_t sephetho = {0};
EI_IMPULSE_ERROR res = run_classifier(&signal, &sephetho,
debug_nn);
haeba (res != EI_IMPULSE_OK) khutla;
bakeng sa (size_t ix = 0; ix < EI_CLASSIFIER_LABEL_COUNT; ix++) {
ei_printf(“%s: %.3f “, result.classification[ix].label,
sephetho.karolo[ix].boleng);
}
#haeba EI_CLASSIFIER_HAS_ANOMALY == 1
ei_printf(“anomaly: %.3f”, result.anomaly);
#endif
ei_printf(“\n”);
}
Sephetho exampLe:
Malebela:
Boloka EI_CLASSIFIER_INTERVAL_MS e hokahane le maqhubu a motho ea fetisang lintlha (mohlala, 100 Hz → 10 ms). Laeborari ea Edge Impulse e beha taba ena ka bo eona ka bo eona ho tsoa ho tšusumetso ea hau.
Haeba o batla ho fumanwa kamehla (fensetere e thellang), qala ho Continuous exampe kenyellelitsoe le laeborari ea EI le ho fapanyetsana ho bala ho ADXL345.
Re tla eketsa lithuto tsa video haufinyane; ho fihlela ka nako eo, lula u mametse - https://www.youtube.com/@RobuInlabs
'Me haeba u ntse u e-na le lipelaelo, u ka sheba video ena ka Edged Impulse: https://www.youtube.com/watch?v=FseGCn-oBA0&t=468s
Litokomane / Lisebelisoa
 |
Arduino ABX00087 Boto ea Nts'etsopele ea WiFi ea UNO R4 [pdf] Bukana ea Mosebelisi R4 WiFi, ADXL345, ABX00087 UNO R4 WiFi Development Board, ABX00087, UNO R4 WiFi Development Board, Boto ea Ntlafatso ea WiFi, Boto ea Nts'etsopele, Boto. |